Evaluation of Electrolyzed Oxidizing Water Solutions

The Pennsylvania State University

The Graduate School

Department of Agricultural and Biological Engineering

EVALUATION OF ELECTROLYZED OXIDIZING WATER

SOLUTIONS AS ALTERNATIVES FOR MILKING SYSTEM

CLEANING-IN-PLACE AND THE DEVELOPMENT OF

MATHEMATICAL MODELS

A Dissertation in

Agricultural and Biological Engineering

Xinmiao Wang

 2015 Xinmiao Wang

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

May 2015 The dissertation of Xinmiao Wang was reviewed and approved* by the following:

Ali Demirci Professor of Agricultural and Biological Engineering Dissertation Co-advisor Chair of Committee

Virendra M. Puri Distinguished Professor of Agricultural and Biological Engineering Dissertation Co-advisor

Paul H. Heinemann Professor of Agricultural and Biological Engineering Head of the Department of Agricultural and Biological Engineering

Robert F. Roberts Professor of Food Science

Robert E. Graves Professor Emeritus of Agricultural and Biological Engineering Special Member

*Signatures are on file in the Graduate School

ABSTRACT

Cleaning and sanitizing of the food processing equipment are important and essential for the safety of food products. Specifically, the effective cleaning and sanitizing of milking system is essential to ensure the dairy product quality and safety. To achieve that, the cleaning and sanitizing of milking system on a dairy farm after the milking event are completed using a highly automated procedure referred as “cleaning-in-place (CIP)”. The chemicals used in the milking system CIP, however, are potentially hazardous to the farmers and the environment. Therefore, novel approaches are needed to solve this problem and further investigate the approaches to optimize the CIP process including investigation of the mechanism of milking system CIP.

Electrolyzed oxidizing (EO) water is an emerging technology for cleaning processing systems in the food industry and as a cleaning and disinfecting agent. By electrolyzing dilute sodium hydroxide solution within an electrolytic chamber separated by a membrane, acidic EO water and alkaline EO water are generated simultaneously. Previous studies in our lab had already demonstrated the efficacy of using EO water as a potential alternative for the milking system CIP on a lab scale pilot milking system. Using EO water is cost effective once the EO water generator is purchased. Most importantly, it is environmentally benign as compared to the concentrated hazardous typical CIP chemicals.

Building on past studies, this dissertation focused on further evaluating the application of

EO water through a real world farm trial and conduct the operational cost comparison between using EO water and conventional chemicals for the milking system CIP. Additionally, another potential alternative to apply EO water was investigated, by combining alkaline and acidic EO water to formulate a blended EO water solution to conduct a one-step CIP on a lab scale pilot milking system. Moreover, a systematic study was carried out of the raw milk deposit removal mechanism during the alkaline and acidic EO water CIP and the optimized blended EO water

iii one-step CIP process by mathematically developing deposit removal rate models for each CIP cycle.

In this study, the real world validation of applying EO water on a commercial dairy farm was conducted and the operational cost was compared with that of using the conventional CIP.

Results showed that using EO water CIP was performance wise comparable to using the conventional CIP but with a lower operational cost (25%). Additionally, an optimization process was performed to determine the optimal condition for the blended EO water one-step CIP on a lab scale pilot milking system. The optimal condition was found to be a cleaning time of 17 min, a starting temperature of 59°C and an acidic EO water percentage of 60% in the blended EO water solution to achieve a 100% CIP performance. When comparing the CIP performance of using the optimized blended EO water and the commercially available one-step CIP chemicals, results showed that using the optimal blended EO water solution was as good as the commercially available one-step chemicals and the operational cost of using the optimal blended EO water CIP was only about 20% as compared to the commercial one-step CIP. Moreover, the contribution of each CIP cycle to the entire milking system CIP process was studied on a stainless steel surface evaluation simulator by developing deposit weight based mathematical mechanistic models. A set of raw milk deposit removal kinetic models were developed and validated firstly for the alkaline and acid wash CIP process. Results showed a substantial amount (more than 90% of the initial deposited soil) of deposit removal occurrence during the initial warm water rinse cycle and a two- stage zeroth order deposit removal during the alkaline wash and acid wash cycles. Based on the proposed models, a 55% reduction of the original CIP operational time was achieved, and this is essentially important from an energy saving aspect. To further explore the raw milk deposit removal mechanisms under different CIP conditions, similar experiments were also conducted for the optimal blended EO water one-step CIP and the corresponding models were developed and validated. Moreover, scanning electron microscopy (SEM) was used to examine the remaining

iv deposit morphology visually on the specimen inner surface. The SEM micrographs served to better understand and explain the deposit removal process during CIP.

In conclusion, this study validated the potential application of alkaline and acidic EO water on commercial dairy farms and investigated another possible one-step CIP method on a lab scale pilot milking system with success. Additionally, from the developed raw milk deposit removal models for the CIP procedures, a shortened operational CIP time duration was achieved.

It was found that the raw milk deposit removal behaved differently under differently CIP processes, and the contribution of the initial warm water rinse cycle is of great significance in the raw milk deposit removal. Therefore, it is concluded that EO water can be applied as alternative milking system CIP solutions. This project made an effort to answer some fundamental questions for dairy farmers and industries to make an informed decision.

TABLE OF CONTENTS

LIST OF FIGURES ...... x LIST OF TABLES ...... xiii ACKNOWLEDGEMENTS ...... xiv TECHNICAL ACKNOWLEDGEMENTS ...... xv

1. INTRODUCTION ...... 1 2. LITERATURE REVIEW ...... 6 2.1. Biofilm and biofouling formations on milk and dairy processing equipment ...... 9 2.1.1. Biofilm formation in dairy ...... 9 2.1.2. Biofouling formation on dairy processing equipment ...... 13 2.1.3. Intrinsic and extrinsic factors affecting biofilm formation on dairy processing equipment ...... 13 2.1.3.1. Species and strain differences ...... 14 2.1.3.2. Temperature of processing conditions ...... 14 2.1.3.3. Flow characteristics ...... 15 2.1.3.4. pH ...... 16 2.1.3.5. Presence of nutrients ...... 17 2.1.4. Control strategies of biofilm on dairy processing equipment ...... 18 2.1.4.1. Prevention and disruption of biofilm and biofouling ...... 18 2.1.4.2. Removal of biofilm with addition of surfactant or enzymes ...... 19 2.1.4.3. Removal of biofilm with modified contact surfaces ...... 20 2.1.4.4. Cleaning methods ...... 21 2.1.5. Monitoring and detection of biofilm in dairy processing equipment...... 23 2.2. Cleaning-in-place for milking system ...... 27 2.3. Electrolyzed oxidizing (EO) water ...... 32 2.3.1. Advantages and disadvantages of EO water ...... 35 2.3.2. EO Water used as a disinfectant in the food industry ...... 36 2.3.2.1. Acidic EO water ...... 36 2.3.2.2. Near-neutral EO water ...... 38 2.3.2.3. Other combinations ...... 39 2.4. Using EO water as an alternative CIP approach for milking system ...... 40 2.5. Milk fouling and CIP process models ...... 42 2.5.1. Milking fouling mechanisms ...... 43

2.5.2. Deposit removal models ...... 45 2.5.3. Surface characterization methods ...... 51 2.6. Summary of literature review ...... 54 3. ELECTROLYZED OXIDIZING WATER FOR CLEANING-IN-PLACE OF ON-FARM MILKING SYSTEMS – PERFORMANCE EVALUATION AND ASSESSMENT ...... 57 3.1. Abstract ...... 57 3.2. Introduction ...... 58 3.3. Materials and methods...... 61 3.3.1. Preparation of the EO water ...... 61 3.3.2. Farm trial ...... 63 3.3.3. Sampling sites ...... 65 3.3.4. Sampling protocols ...... 66 3.3.5. Sample analyses ...... 67 3.3.6. Cost analyses ...... 68 3.4. Results and discussions ...... 71 3.4.1. ATP RLU readings ...... 71 3.4.2. Microbiological enrichment ...... 76 3.4.3. Cost comparison ...... 79 3.5. Conclusions ...... 81 3.6. Acknowledgements ...... 82 3.7. References ...... 82 4. EVALUATION OF BLENDED ELECTROLYZED OXIDIZING WATER-BASED CLEANING-IN-PLACE TECHNIQUE USING A LAB SCALE PILOT MILKING SYSTEM ...... 85 4.1. Abstract ...... 85 4.2. Introduction ...... 86 4.3. Materials and methods...... 88 4.3.1. Microorganisms and medium ...... 88 4.3.2. Description of the pilot milking system for one-step cleaning trial ...... 89 4.3.3. Preparation of blended EO water solutions ...... 90 4.3.4. Preparation of the milking system for CIP process...... 91 4.3.4.1. Shock cleaning ...... 91 4.3.4.2. Soiling the system ...... 92 4.3.5. Experimental design ...... 94 4.3.6. Evaluation ...... 95

vii

4.3.7. Analysis ...... 97 4.3.7.1. ATP bioluminescence ...... 97 4.3.7.2. Calcium alginate sampling ...... 98 4.3.7.3. Operational cost analysis...... 98 4.3.7.4. Statistical analysis ...... 99 4.4. Results and discussions ...... 99 4.4.1. Determining the range of parameters evaluated for one-step CIP ...... 100 4.4.2. Optimization of the blended EO water CIP process ...... 103 4.4.3. CIP performance comparison between the optimal blended EO water and the commercial one-step chemicals ...... 109 4.4.4. Operational cost comparison between the optimal blended EO water CIP and the commercial one-step CIP ...... 112 4.5. Conclusions ...... 114 4.6. Acknowledgements ...... 115 4.7. References ...... 115 5. MATHEMATICAL MODELING AND CYCLE TIME REDUCTION OF DEPOSIT REMOVAL FROM STAINLESS STEEL PIPELINE DURING CLEANING-IN-PLACE OF MILKING SYSTEM WITH ELECTROLYZED OXIDIZING WATER ...... 118 5.1. Abstract ...... 118 5.2. Introduction ...... 119 5.3. Materials and methods...... 122 5.3.1. Stainless steel surface evaluation by using the simulator ...... 122 5.3.2. Experimental procedure ...... 125 5.3.2.1. Number of raw milk soiling ...... 125 5.3.2.2. Drying temperature and time and cooling protocol ...... 126 5.3.2.3. Description of one complete set of experiments ...... 128 5.3.2.4. Use of nondimensionalized data ...... 130 5.3.2.5. Mathematical model development ...... 131 5.3.2.6. Alternative ATP bioluminescence method ...... 141 5.3.2.7. Visual evaluation of the specimen inner surface morphology ...... 142 5.3.2.8. Statistical analysis ...... 142 5.4. Results and discussions ...... 142 5.4.1. Statistical comparison ...... 143 5.4.2. Mathematical model development ...... 144 5.4.2.1. Warm water rinse cycle ...... 144 5.4.2.2. Alkaline wash cycle ...... 153 viii

5.4.2.3. Acid wash cycle ...... 156 5.4.2.4. Overall model summary ...... 157 5.4.3. AT P bioluminescence validation ...... 161 5.4.3.1. Original CIP process validation ...... 162 5.4.3.2. CIP process validation ...... 164 5.5. Conclusions ...... 170 5.6. Acknowledgements ...... 171 5.7. References ...... 171 6. ONE-STEP CLEANING-IN-PLACE FOR MILKING SYSTEMS AND MATHEMATICAL MODELING FOR DEPOSIT REMOVAL FROM STAINLESS STEEL PIPELINE USING BLENDED ELECTROLYZED OXIDIZING WATER ...... 174 6.1. Abstract ...... 174 6.2. Introduction ...... 175 6.3. Materials and methods...... 178 6.3.1. Stainless steel surface evaluation by using the simulator ...... 178 6.3.2. Experimental procedure ...... 178 6.3.3. Mathematical model development ...... 179 6.3.4. Alternative ATP bioluminescence method ...... 185 6.3.5. Qualitative evaluation of the specimen inner surface morphology...... 185 6.3.6. Statistical analysis ...... 186 6.4. Results and discussions ...... 186 6.4.1. Statistical comparisons ...... 186 6.4.2. Mathematical model development ...... 187 6.4.2.1. Warm water rinse cycle ...... 187 6.4.2.2. Blended EO water one-step wash cycle ...... 190 6.4.2.3. Overall model summary ...... 193 6.4.3. ATP bioluminescence validation ...... 194 6.5. Conclusions ...... 196 6.6. Acknowledgements ...... 197 6.7. References ...... 197 7. CONCLUSIONS AND RECOMMENDATIONS ...... 200 REFERENCES ...... 206

LIST OF FIGURES Figure 2-1. Stages of biofilm development ...... 10 Figure 2-2. Tie-stall barn milking system at a commercial dairy farm ...... 30 Figure 2-3. Schematic of electrolyzed water generator and produced compounds at both electrode ...... 33 Figure 2-4. EO water generator ...... 34 Figure 2-5. Different configurations of EO water generator ...... 35 Figure 2-6. Schematic of the previously used pilot-scale milking system ...... 41 Figure 2-7. Description of hypothesized fouling model involving the interactions of different stages of proteins in the bulk fluid, in the thermal boundary layer and on the contact surface ...... 44 Figure 2-8. Schematic of the stages involved in removal of whey protein deposits ...... 45 Figure 2-9. SEM images of stainless steel 316 surfaces with different surface finishes of 2B and 2R ...... 49 Figure 2-10. Deposition on 2R finish surfaces from modified simulated milk ultrafiltrate solution ...... 54 Figure 2-11. SEM micrographs of the mostly amorphous deposits formed from simulated milk ultrafiltrate at pH = 6.3 and 60°C (flow velocity 0.32 m/s) at various run times on 2R finish surfaces ...... 54 Figure 3-1. Mechanism of the EO water generation ...... 59 Figure 3-2. Overall schematic of the EO water generator and other units for milking system CIP on a commercial dairy farm and associated flow loops ...... 62 Figure 3-3. (a) General overview schematic of farm milking and cleaning system; (b) detailed sampling locations along the pipelines in the milk house ...... 65 Figure 3-4. Sampling schematic detail ...... 67 Figure 3-5. Average RLU reading comparison between the EO water CIP and the conventional CIP ...... 72 Figure 3-6. Negative bacterial enrichment comparison between the EO water CIP and the conventional CIP ...... 77 Figure 4-1. Pilot plant milking system and sampling location schematic of the pipes and elbows ...... 90 Figure 4-2. Location sampling protocols of the pipes, elbows and other milking system components ...... 96 Figure 4-3. RLU reduction percentages at different acidic EO water percentages for sampling categories of pipes, elbows and other milking system components of gaskets, liners, and milk hose ...... 101 Figure 4-4. Negative enrichment percentages at different acidic EO water percentages for sampling categories of pipes, elbows and other milking system components of gaskets, liners, and milk hose ...... 102 Figure 4-5. 3D surface plots of RLU reduction percentage for sampling locations of pipes ...... 107 Figure 4-6. 3D surface plots of RLU reduction percentage for sampling locations of elbows ...... 107

Figure 4-7. CIP performance comparison among two commercial one-step cleaning chemicals and blended EO water at its optimal conditions ...... 108 Figure 5-1. Stainless steel surface evaluation simulator with specimen of 152.4 mm straight pipe test section...... 124 Figure 5-2. A 152.4 mm stainless steel straight pipe specimen in the simulator and recovery section ...... 125 Figure 5-3. Deposit mass on the inner surfaces of upstream located specimens #1, 2 and 3 after different drying time at 70°C in the incubator ...... 127 Figure 5-4. Illustration of the proposed two-term deposit removal model during the warm water rinse cycle ...... 135 Figure 5-5. Illustrations of proposed two-term zeroth power proportional model during the alkaline wash cycle and acid wash cycle ...... 137 Figure 5-6. Specimen inner surface morphology at the end CIP cycles ...... 145 Figure 5-7. Nondimensionalized experimental deposit weight change during the CIP process with the proposed model for the upstream locations ...... 147 Figure 5-8. Interval plot of nondimensionalized experimental deposit weight change during the CIP process with 95% confidence interval for upstream locations ...... 149 Figure 5-9. Experimentally calculated deposit removal rate and developed model proposed deposit removal rate for upstream locations ...... 151 Figure 5-10. Interval plot of nondimensionalized experimentally calculated deposit removal rate during the CIP process with 95% confidence interval for upstream locations ...... 152 Figure 5-11. Nondimensionalized deposit mass and log(RLU+1) at the end of the original CIP cycles of 30 s warm water rinse, 10 min alkaline and 10 min acid wash ...... 163 Figure 5-12. Nondimensionalized deposit mass and log(RLU+1) at the end of the optimized CIP cycles of 10 s warm water rinse, 3 min alkaline and acid wash...... 166 Figure 5-13. Nondimensionalized deposit mass and log(RLU+1) at the end of the further optimized CIP cycles of 10 s warm water rinse, 3 min alkaline and 6 min acid wash ...... 167 Figure 5-14. Linear relationship between the nondimensionalized deposit mass and the log(RLU+1) at the end of the further optimized CIP cycles of 10 s warm water rinse, 3 min alkaline and 6 min acid wash, without the initial nondimensionalized deposited mass and initial ATP RLU reading ...... 169 Figure 6-1. Illustration of the proposed two-term deposit removal model during the warm water rinse cycle ...... 182 Figure 6-2. Illustration of the proposed two-term deposit removal model during the blended EO water one-step wash ...... 184 Figure 6-3. Typical inner surface morphology for a specimen at the end of different stages of CIP cycle ...... 188 Figure 6-4. Nondimensionalized experimental decrease in deposit weight during the warm water rinse cycle comparison with the proposed model ...... 189 Figure 6-5. Nondimensionalized experimental decrease in deposit weight during the CIP process comparison with the proposed model ...... 191

Figure 6-6. Nondimensionalized deposit mass and log(RLU+1) of at the sampling time points of blended EO water one-step CIP and comparison with that of at the end of the previously optimized 10 s warm water rinse/3 min alkaline wash/6 min acid wash CIP result ...... 196

xii

LIST OF TABLES Table 3-1. CIP Recommendations from Dairy Practice Council ...... 59 Table 3-2. Cleaning condition comparisons per cycle between the EO water CIP and the conventional CIP on a commercial dairy farm ...... 64 Table 3-3. Operational cost comparison between the EO water and conventional CIP on a commercial dairy farm ...... 71 Table 3-4. p-values between the conventional CIP and EO water CIP ATP bioluminescence evaluations for the sampling locations of straight pipelines, elbows, gaskets, liners, milkhoses, and milk inlets ...... 76 Table 4-1. EO water solution property comparisons ...... 91 Table 4-2. Operation process comparison between the blended EO water and commercial one- step chemicals for pilot milking system CIP ...... 93 Table 4-3. Three factor Box-Behnken experimental design for CIP optimization process ...... 94 Table 4-4. Response surface method result for each sampling category ...... 104 Table 4-5. CIP parameter comparisons between using two different commercial one-step chemicals and the blended EO water at its optimal condition ...... 110 Table 4-6. Operational cost comparison between the blended EO water and commercial one-step CIP on the pilot milking system ...... 114 Table 5-1. Sampling time points for the upstream and downstream located specimens during the CIP cycles ...... 130 Table 5-2. Coefficient of variance of the nondimensionalized experimental deposit weight per unit contact area (mg/mg/m2) during the CIP process...... 150 Table 5-3. Coefficient of variance of the nondimensionalized experimentally calculated deposit removal rate (mg/mg/m2/s) during the CIP process ...... 153 Table 5-4. Percent error difference during different CIP cycles for the upstream and downstream locations ...... 159 Table 5-5. Percent variation during different CIP cycles for the upstream and downstream locations ...... 161

xiii

ACKNOWLEDGEMENTS

I would like to express my heartfelt gratitude to my advisors, Dr. Ali Demirci and Dr. Virendra Puri, who helped me along my Ph. D. study. I wouldn’t be able to achieve any of this without their help. They not only guided me for my Ph. D. studies; but also taught me heavily on how to be an excellent researcher, educator, team worker and more. They are the role models that I am working toward to, and I wish I could be like them some day in my career. I want to thank Dr. Robert Graves, who guided me along the way with so much help. I am wordless to express my gratitude. He devoted so much time on my project and asked for nothing in return. Without his help, I wouldn’t be able construct the CIP similar system. I learned a lot from him – how to be a questioner, an observer, a problem solver, and a good communicator. Thank you. I would like to also express my appreciation to Dr. Paul Heinemann and Dr. Robert Roberts, who served on my committee and provided constructive suggestions. Both of them are department heads with loaded schedule, while they still spare their time for me and I am so grateful for that. Their academic support and scientific insight are greatly appreciated. Thank you. I would like to thank Dr. Ernest Hovingh and Dr. Steve Spencer, who helped me greatly when I was setting up the milking system in the lab; I will always remember their expertise and great help. Gratitude goes to my lifetime mentor Dr. Dechang Xu, who is more than an advisor, but an integral role model; and my teachers, Mingfu Wu, Xiquan Xia, Jiashun Ding and Ruirui Du in China. Additionally, I would like to thank all the Department of Agricultural and Biological Engineering members, who made me feel a part of their warm family. Special thanks go to Dr. Roderick Thomas, Mr. Randall Bock and Mr. Bill Harshman, who helped me many times when I was in trouble feeling desperate. I hope to visit again and I wish everyone health, fortune and happiness. My gratitude also goes to all personnel at Penn State dairy barn. In the last year of my Ph. D. study, you helped me greatly. I couldn't even imagine all these five o’clock early mornings without you and I would like to deliver my sincerest appreciation and I wish all of you happiness. I am thankful to have met my lab mates – Dr. Duygu Ercan, Dr. Hasan Coban, soon-to-be Dr. Gulten Izmirlioglu, Ozge Can, Dr. Gulsad Uslu, and Ehsan Mahdinia, who helped me in so many ways on so many days. I wish all of you best and fortune. Additionally, I want to thank Dr. Ann Kusnadi, who is such a wonderful person and inspired me in so many ways; I couldn't even be the person I am today without her helpful guidance, interesting jokes, persistent pursuit and of course, amazingly delicious fine cuisine. I am also thankful that I had encountered with so many endearing friends, who helped me one way or another. Special thanks go to Xiaoxue Wang, for her hearty support; Zhao Zhang, for all these summer mornings’ backup and fortune talks; Wenqing Yao and Dan Hofstetter, for their on-call technical help; Juan Tao, for her warm support; Yang Xing, Yingnan Shi, Yunfei Zhang, Apoorva Karamchandani, Zhaoran Li, Boya Xiong, Tianyi Wang, Xiwen Zhang and Ye Cao, Shiming Lei, Fei Xue, Justin Wen, and all others that might just slip off my lip. Last, but definitely not the least, I would like to express my gratitude to my family members. My siblings, Yin Liu, Yingrui Liu, Zihui Liu, Zetian Qu, and Nan Zhang; uncles and aunts, Yunlong Liu and Fuzhen Wang, Yunxing Liu and Zhaofen Ding, Yunyan Liu and Yuling Liu, Yunxia Liu and Daifeng Qu, Ling Wang and Jinsheng Zhang, Dawei Wang, Qi Wang and Yan Wang; and more importantly, my parents, Ke Wang and Yuying Liu. Thank you for encouraging me to follow my dreams. I love you all.

xiv

TECHNICAL ACKNOWLEDGEMENTS

This work was supported by the USDA Special Research Grant (No. 2010-34163-21179) and Pennsylvania Agricultural Experiment Station. I would also like to thank Hoshizaki Electric

Co. Ltd (Sakae, Toyoake, Aichi, Japan) for the technical help, Fisher and Thompson (Belleville,

PA) for the milking system set-up and trouble-shooting help, and Charm Sciences (Lawrence,

MA) for their technical help on ATP bioluminescence test.

CHAPTER 1

INTRODUCTION

The most recent data from the Food and Agriculture Organization (FAO) of the United

Nations indicated that United States ranked first in the world for the production of whole, fresh cow milk (FAOSTAT, 2012). It is therefore of great importance to guarantee the quality and safety of milk from the very beginning of the process line to the very end point of the marketplace. In the dairy industry, processing equipment surfaces are easily contaminated by biofilm-forming microorganisms and if handled inappropriately, it would affect the product quality and the process efficiency. The biofilm-forming microorganisms are bacteria, yeasts, or molds, and the adherent surfaces include stainless steel, glass, plastic, rubber, etc. Term biofouling is often seen in literature accompanied with biofilm; it describes not only the film structure that microorganisms form, but also the organic matter trapped from biofilm surroundings and the deposits formed together (Vlkova and Babak, 2008). Biofouling is often observed along with biofilm in food processing equipment due to the natural composition of food including the organic and inorganic constituents. Similarly, biofouling is a major concern in dairy industry (Choi et al., 2013).

In the dairy industry, the handling and processing of milk is subject to strict regulations.

In the United States, these regulations cover from the planning of a dairy barn and the selection of equipment, to the cleaning and sanitizing of all the transportation and processing equipment and systems and the testing procedures of milk products (DPC, 2010). Every dairy farm, small or large, must comply with the requirements and is subject to inspection on a regular basis.

Cleaning-in-place (CIP) is a process in which the equipment is cleaned and sanitized without disassembly. CIP is especially useful in the food industry due to its high automation and low

1 downtime. The quality of CIP affects the processed product; therefore special attention must be paid to the performance of the CIP process.

Milking system CIP usually starts right after the completion of a milking event, and conventionally, it consists of four steps (DPC, 2010). The first step is a warm water rinse cycle.

40℃ water is used in this cycle to remove the residual deposit from milk contact surfaces. The second step is an alkaline wash cycle. The wash solution has a starting wash solution temperature around 70℃ in order to remove the organic deposit including the milk fat and protein. The acid wash cycle follows the alkaline wash cycle. The acidic wash solution helps to remove the inorganic deposit of residual minerals and inhibit microbial growth. Right before the next milking event, there is a sanitizing circulation. The solution used during the sanitizing circulation is chlorinated. Typically, these CIP cycles use costly and concentrated chemicals, and they possess potential hazard to the farmers and the environment. Therefore, there is a need to investigate cheaper and safer alternatives, which are also environmentally benign to substitute the harsh CIP chemicals.

Electrolyzed oxidizing (EO) water is a novel cleaning and sanitizing agent, which has been studied widely in recent years. Alkaline and acidic EO water solutions are generated simultaneously by electrolyzing a weak salt solution in an electrified chamber, separated by a membrane between the positively and negatively charged electrodes. Under certain conditions, the pH of the alkaline EO can be as high as 11.5 with an oxidation-reduction potential (ORP) of -

795 mV, and the pH of the acidic EO can be as low as 2.6 with an ORP of 1,150 mV and a residual chlorine of about 80 ppm (Huang et al., 2010). EO water has been evaluated for decontamination of many food products successfully. Acidic EO water has been evaluated for the decontamination on the fresh pork surfaces (Fabrizio et al., 2005), inoculated egg surfaces (Bialka et al., 2004) or hatching eggs (Fasenko et al., 2009). Moreover, inoculated salmon fillets were treated with alkaline EO water followed by acidic EO water (Ozer & Demirci, 2006). On the

2 other hand, studies in our lab examined the possibility of using the alkaline and acidic EO water solutions as an alternative for milking system CIP. The performance of EO water CIP had been demonstrated using milking system related materials including stainless steel, plastic and rubber

(Walker et al., 2005a) as well as on a lab scale pilot milking system. Dev et al. (2014) further optimized the factors affecting the EO water CIP of the wash solution temperature. However, no studies have been conducted under a real-world scenario, which is crucial to validate the EO water CIP performance on a commercial dairy farm to complete the study series.

Additionally, EO water generators are available, which produce only one EO water solution named as “near-neutral EO water”. In the production of the near neutral EO water, the original separation membrane was removed from the electrodialysis chamber; therefore there is a simultaneous mixing/blending process during the electrodialysis. This process results in a less corrosive blended EO water solution and study showed a satisfactory sanitizing capability using this solution (Guentzel et al., 2008). Coincidentally, a new CIP approach has received attention on increasing numbers of dairy farms (Parr, 2013), which combines the alkaline wash and the acid wash cycles into one as a “one-step” CIP for the milking system. Several commercially available one-step CIP chemicals are on the market and their CIP performance is claimed to be comparable to the conventional separate alkaline and acid wash CIP. This one-step CIP has a clear advantage when compared to the conventional CIP, because it potentially costs less due to the elimination of one CIP cycle, plus the reduction in the water, chemical, energy, and time usage. Based on the published studies involving using EO water for the milking system CIP and the characteristics of the near-neutral EO water, it is worth studying the blending process of the alkaline and acidic EO water solutions and applying that on a lab-scale pilot milking system as an alternative for the one- step CIP.

Furthermore, studies are needed for the raw milk deposit removal mechanism during the

CIP process, to be specific, the detailed contribution of each CIP cycle to the entire CIP process.

This could be accomplished through the development of a mechanism-based deposit removal rate mathematical model. To achieve this, a stainless steel surface evaluation simulator is needed to conduct the experimental and validation runs. The deposit removal rate models study the soiled raw milk deposit removed during the warm water rinse cycle, the alkaline wash cycle, and the acid wash cycle. In addition, given the progress made for the blended EO water as one-step milking system CIP, similar studies could also be expanded to the development of the deposit removal rate model for the one-step CIP. Comparisons could be made among these models from the perspective of deposit residual weight change, the deposit removal rate change, and the specimen inner surface deposit morphology change after each cleaning cycle during the CIP process.

The utilization of EO water for milking system CIP would bring benefits to the farmers, the society, and the environment. However, more work is needed to further promote this technology, demonstrate its effectiveness under different scenarios, and study the deposit removal mechanism during the CIP process. This study aims to address some fundamental issues that are hurdles to further development of using and acceptance of the EO water technology for the milking system CIP and the answers from this research could bring direct benefits for farms and farmers of today and beyond.

In this dissertation, Chapter 1 presents the overall introduction to the dissertation. Chapter

2 summarizes in detail the literatures related to this dissertation along with some expanded information which might be useful for the readers. Chapter 3 presents the evaluation of applying the alkaline and acidic EO water solutions in the real world on a commercial dairy farm as an alternative CIP method, and the CIP performance and cost comparisons between using EO water solutions with that of using conventional chemicals are conducted. Chapter 4 deals with the study of using different blending combinations of the alkaline and acidic EO water as the one-step milking system CIP methods and achieve an optimal one-step CIP condition. During the

4 optimization, solution starting temperature and circulation time are also taken into account as optimizing factors. Chapter 5 presents a more detailed study of using a stainless steel surface evaluation simulator for the development of a milk deposit removal rate model for alkaline and acidic EO water CIP process. An optimized CIP process is also proposed based on the developed model and validated using ATP bioluminescence method. Chapter 6 presents the deposit removal rate model development of optimal blended EO water as a one-step CIP process and concludes with the specimen inner surface morphology comparison of different CIP processes under different CIP cycles. Chapter 7 summarizes the outcome of this study and also presents potential future research topics to focus on.

CHAPTER 2

LITERATURE REVIEW

According to the latest data from the Food and Agriculture Organization of the United

Nations, the United States produced about 90 million metric tons of cow milk, ranking first in the world (FAOSTAT, 2012). The annual average consumption of dairy products per capita in the

United States has increased from 244.5 kg in 1975 to 277.6 kg in 2012 (USDA, 2012). The Food and Drug Administration has issued several regulations to regulate the raw milk quality in the

United States (HHS/FDA, 2007). Therefore, it is of great importance to ensure the quality of raw milk and other dairy products and provide the public with safe products to consume.

Due to the high nutritional value of milk and dairy products, raw milk may harbor several pathogenic microorganisms such as Listeria, Salmonella, Escherichia coli O157:H7, and

Campylobacter jejuni, which can be harmful to some people. These and other microorganisms have caused outbreaks in the past years in the United States (FDA, 2012). A most recent case occurred on a dairy farm in Tennessee in November 2013, in which the existence of E. coli

O157:H7 in raw milk caused nine young children illness after consumption; five of whom were hospitalized, and three developed a severe kidney problem known as hemolytic uremic syndrome

(Beecher, 2013). Another outbreak occurred in Pennsylvania in May of 2013, when the

Pennsylvania Department of Health reported the presence of C. jejuni from raw milk. A probable diarrheal illness was diagnosed from a person, who consumed the raw milk from this

Pennsylvania farm and four other cases involved children whose ages were under 18 (Weltman et al., 2013). For pasteurized milk, Streptococci and the endospore forming Bacillus spp. are also a major concern; both are capable of biofilm formation. All of these cases indicate the significance of maintaining a cleaned and sanitized milk contact surfaces.

In the dairy processing systems, biofilms are contaminating microorganisms, and these microorganisms adhere to solid surfaces in the facility and produce simple or complex structures.

Similarly, in the food industry, the fluid processing equipment surfaces are commonly contaminated by these biofilm-forming microorganisms, which affect the food product quality and the process efficiency. The microorganisms are bacteria, yeasts, or molds, and the adherent surfaces vary from food contact surfaces like stainless steel and glass to common surfaces such as floors and conveyer belts.

Biofouling, often seen in literature accompanied with biofilm, describes not only the film structure that microorganisms form, but also the organic matter trapped from biofilm surroundings and the deposits formed together (Vlkova and Babak, 2008). Biofouling is often observed along with biofilm in food processing equipment due to the food natural composition with organic and inorganic composition. Especially, biofouling is a major concern in dairy industry from the nutritious constituents in milk (Choi et al., 2013).

For the dairy industry, the handling and processing of milk is subject to strict regulations to prevent any potential outbreak. In the United States, these regulations cover from the very beginning of planning a dairy barn such as the installation and the structure, to the equipment recommendation, and the cleaning and sanitizing of all the transportation and processing equipment and the testing of milk and the final dairy product (DPC, 2010). Every dairy farm, small or large, must comply with the requirements and conduct inspections on a regular basis.

Dairy processing equipment possesses a variety of milk contact surfaces, which may become potential microbial habitat if the cleaning and sanitizing process is handled inappropriately, or the equipment is not hygienically designed. The materials and configuration of these contact surfaces varies. Some common materials or surfaces are glass, stainless steel, plastic, and elastomers (or rubber) and different types of membrane used for filtration processes.

Since the treatment parameters are different for each process, the cleaning and sanitizing method 7 varies. The different cleaning and sanitizing procedures make it difficult to unify all methods into a universal method.

Cleaning and sanitizing of the food processing lines are important and essential for the safety of the food products. There are several cleaning agents and sanitizers commonly used in the food industry, such as alkaline, chlorine compounds, organic acids, trisodium phosphate, iodophores, and quaternary ammonium compounds (Hricova et al., 2008). These chemicals are costly, and might bring potential residual chemical hazard to the system and the environment.

Automated Cleaning-In-Place (CIP) is able to reach a cleaner state compared with other cleaning methods (Tamime, 2008). A considerable attention was put into the study of milking system CIP due to the complexity of the milking system and the importance to keep the system clean and sanitized properly. Electrolyzed oxidizing (EO) water is a novel technology, which has been proven to be effective in cleaning and sanitizing. Literature review related to this study is divided into the following categories; i) Biofilm formation in milk and dairy processing equipment and the monitoring and control strategies are explicitly explained; ii) CIP approach and its utilization in the conventional milking system cleaning are introduced; iii) Detailed information about EO water is provided including the development history, covered studies, and to be specific, near- neutral EO water is introduced; iv) Previous studies of using EO water as an alternative approach for milking system CIP are discussed; v) Studies of milk fouling and CIP models including the fouling mechanisms, the deposit removal models during CIP, the advanced computational fluid dynamics models, and the characterization methods commonly used in milk contact surfaces are introduced; and vi) the summary of literature review.

2.1 Biofilm and biofouling formations on milk and dairy processing equipment

2.1.1 Biofilm formation in dairy

In nature, microorganisms are found mainly attached to surfaces as a survival strategy to overcome adverse environmental changes such as pH, temperature, and substrate (Stepanović et al., 2004; Stepanović et al., 2003). Generally, the formation of biofilm includes the following stages: (1) the initial attachment, (2) firmly irreversible attachment, (3) early development of the bioflim architecture, (4) maturation of the biofilm architecture, and (5) dispersion of single cells from the bioflim (fig. 2-1).

The initial attachment of the microorganisms is basically the cell adhesion onto the contacting surfaces. This process strongly depends on the physiochemical properties of the bacterial cell surface (Ferreira et al., 2010). These initial adherent cells begin to produce a small amount of extracellular polymeric substance (EPS) in the beginning of the attachment (O’Toole and Kolter, 1998). The adhesion in this stage is reversible, since the attached microorganisms have not differentiated yet with morphological changes, which lead to the formation of biofilm; many of these attached cells may detach from the surface and back to be planktonic condition again (Stoodley et al., 2002b). It is important to notice the role of the surface material during this stage. In one study, the researchers showed that compared with hydrophilic surface, Salmonella and Listeria tend to attach more easily to the hydrophobic surfaces (Sinde and Carballo, 2000).

The change from the initial attachment to the irreversible attachment is largely due to the presence of EPS causing a stronger bonding to the surface (Stoodley et al., 2002b). When the irreversible attachment is formed, the removal of biofilm is harder – increased shear forces and chemicals breaking down the linkages are needed from the cleaning and sanitizing fluid (Sinde and Carballo, 2000) and sometimes heat is also needed (Augustin et al., 2004; Sinde and

Carballo, 2000).

Biofilm formation continues to develop after irreversible attachment, since the initial growing condition is satisfied. The continuous accumulation of microorganisms and the production of EPS facilitate the biofilm formation (Chmielewski and Frank, 2003). The already produced EPS, in turn, strengthens the bond between the microorganism and the substratum

(Donlan, 2002). The biofilm maturation is the spatial arrangement of the colonies within the biofilm matrix; this leads to an organized biofilm structure such as a mushroom shape or a flat shape (Chmielewski and Frank, 2003). These arrangements are fundamental to the function of the complex biofilm (Davey et al., 2003).

Figure 2-1. Stages of biofilms development, from the initial attachment (1) then the firmly

attachment (2) to the development (3) of biofilm structure and maturation (4), and a

finally partial dispersion (5) of cells from biofilm (Monroe, 2007).

In the last stage, some cells within the biofilms are released from the biofilm and reverted back to the planktonic form (Sauer et al., 2002). There are several approaches to get the biofilm detached from the contacting surfaces, including the mechanical shear forces of the passing fluid

(Stoodley et al., 2002a), enzymatic hydrolysis of EPS and/or surface binding proteins, alkaline

10 solution treatments (Kaplan et al., 2003; Kaplan et al., 2004). Similar to the initial attachment, the lack of nutrient is also a cause for biofilm detachment (O’Toole et al., 2000).

It has been stated that the majority of dairy biofilm consists of the microbial cells and milk residues, most of which are various milk proteins and minerals such as calcium phosphate

(Flint et al., 1997; Mittelman, 1998). One study demonstrated that several types of microorganisms were collected in raw milk line from a commercial plant, such as Bacillus cereus,

Bacillus subtilis, Bacillus spp., Streptococcus spp., Lactococcus spp., Lactobacillus spp., E. coli,

Enterobacter aerogenes, Shigella spp., and some yeasts. There are species differences with respect to sampling locations along the processing lines – some of these microflora were not present at the pasteurizer outlet, while some other microorganisms initially not found in raw milk line were present at the buffer tank outlet (Sharma and Anand, 2002).

Listeria monocytogenes is a major concern in raw milk due to its psychrotrophic nature.

The bacterium is often related with foodborne outbreaks and causes a high mortality rate (Borucki et al., 2003). In addition to the low temperature survival characteristic, a seven-year study showed that a clone of L. monocytogenes most likely survived in a Scandinavian dairy industry plant for, at least, seven years (Unnerstad and Bannerman, 1996). However, these non-spore-forming phychrophic microorganisms are killed by the process of pasteurization; whereas the spore- forming Bacillus species, which exist on the surfaces of milk processing plant, endospores can survive pasteurization (Dat et al., 2012). When raw milk undergoes pasteurization process, chances are some endospores would survive, and are heat shocked into germination; this leads to the formation of biofilm through the adherence to stainless steel, the surface of which is suitable for growth. It is suspected that the endospores stick to the stainless steel by forming a monolayer instead of the traditional conception of multilayer bacterial cells embedded in an EPS matrix with water channels. Another study assessed the air-liquid biofilms of different strains of mesophilic,

B. cereus and discovered that thick and higher amount of biofilms were developed at the air-

11 liquid interface compared with those formed in submerged systems (Wijman et al., 2007). The results indicated a possible phenomenon that B. cereus biofilms might develop particularly in partially-filled storage and piping systems during operation or the remaining liquid level after one production cycle. Once the biofilm is formed, endospores and vegetative cells are embedded inside the biofilm, and they are protected against sanitizers (Ryu and Beuchat, 2005). Studies using B. cereus spore isolated from dairy silo tanks showed that some genotypes were able to survive in hot-alkaline (75°C) wash liquid in the cleaning-in-place (CIP) step and some others endospores even showed well preserved viability when heated at 90°C (Shaheen et al., 2010).

Thermophilic bacteria are another category of microbial dairy product spoiler and biofilm formers. They are able to grow at higher temperatures from 45-65°C. Consequently, they are problematic in a milk powder plant where they are present in the preheating and evaporation processing sections, contaminating the final products (Fratamico et al., 2009). One study demonstrated that the spores of thermophilic Geobacillus spp., which are commonly found in raw milk, are not densely populated; however, a higher population was reported, when sampling on heat exchangers and evaporators (Burgess et al., 2010). Another dominant thermophilic bacterium is Anoxybacillus flavithermus. One study examined the biofilm formation rate of A. flavithermus in skim milk under different temperature settings in a continuous flow test reactor (48-60°C). The researchers enumerated the cell numbers in the milk and on the surface of the stainless steel tubing. Results showed that biofilm formation occurred very rapidly, within 6–8 h after inoculation. The biofilm formation rate reached its peak after 6 h, and in some cases up to 50% of biofilm was found to be endospores after 8 h. However, the temperature variation experiments showed that endospores were formed at 55 and 60°C; not at 48°C. Therefore, the researchers suggested that by lowering the temperature, sporulation could be prevented. In addition, they showed that A. flavithermus, the biofilm formation and sporulation can occur simultaneously compared to other mesophilic dairy contaminants (Burgess et al., 2009).

2.1.2 Biofouling formation on dairy processing equipment

Specific to dairy practices, biofilm often takes place when the processing equipment is subject to thermal treatments. The high processing temperature induces the change in the physiochemical properties of milk then lead to a deposition onto the contact surface. The components of deposition vary in a large range; both organic compounds and minerals are deposited. A typical cow’s milk composition consists of 5% lactose, 3.7% fat, 3.4% protein, and

0.7% minerals which include sodium, potassium, calcium, magnesium, and more (Marth, 1988).

Based on the heating temperature differences, milk fouling is categorized into two: i) between the heating temperature from 75 to 110°C, namely at the pasteurization temperatures, the fouling is primarily protein (takes up to 50-70%); ii) at ultra-high heating temperatures above 120°C, the fouling is mostly mineral content (takes up to 70-80%) (Burton, 1968).

2.1.3 Intrinsic and extrinsic factors affecting biofilm formation on dairy processing

equipment

There are several factors affecting the formation of biofilm on the surface of processing equipment; categorized as intrinsic and extrinsic factors which include species and strains of microorganism, the temperature of the processing conditions, the flow characteristics both chemically and mechanically, the presence of organic and inorganic matters (McLandsborough et al., 2006). Understanding these factors and their function in the biofilm formation would be beneficial in the control and ultimately prevention of the biofilm formation caused by these factors.

2.1.3.1 Species and strain differences

One study compared the adhesion ability of four Salmonella Enteritidis isolates to three different types of materials (Oliveira et al., 2006). In addition to the differences of materials, there were some unexplained results besides the surface hydrophobicity and roughness. They suspected that the adhesion of Salmonella to surfaces is strain dependent – the difference in cell wall protein, fimbriae and flagella contribute to the adhesion process. Another study used ten isolates of L. monocytogenes and eight commonly used kitchen materials, and results demonstrated attachment by all strains and to all surfaces, with the extent of biofilm development quite distinct among trials (Silva et al., 2008). The materials they used included stainless steel, glass, and polypropylene, all of which are commonly seen on dairy farms and dairy processing equipment.

2.1.3.2 Temperature of processing conditions

Temperature affects the growth of microorganisms; therefore, the processing condition temperature affects the biofilm formation. Another study used L. monocytogenes to form biofilm on various food contact surfaces (Di Bonaventura et al., 2008). The researchers tested 44 different strains from various origins of food and environment, and tested the biofilm level under different temperatures and materials. At lower temperatures (4, 12, and 22°C), the biofilm formation levels on glass were higher than those on polystyrene and stainless steel. When temperature increased to

37°C, the biofilm formation levels of glass and stainless steel were significantly higher; however, not much difference was observed for polystyrene. The emergent microorganism Enterobacter sakazakii in infant formula milk was studied to determine its growth, thermotolerance and biofilm formation (Iversen et al., 2004). Researchers conclude that the pathogen was able to grow at refrigerated temperatures and the standard pasteurization process was able to kill it and its biofilm formation.

2.1.3.3 Flow characteristics

A simple approach to distinctly different flow types is the flow velocity (or Reynolds number). One study examined the effect of biofilm formation under turbulent flows in drinking water environment with different materials and roughness (Percival et al., 1999). The flow velocity they used varied from 0.32 to 1.75 m/s. They compared the levels of EPS for each case and observed no significant differences between each flow rate on all stainless steel surfaces

(Stainless steel grade 304 and 316 with smooth and rough surface finishes). However, at higher flow rate, higher levels of EPS were observed – after 5 months of exposure to potable water, on stainless steel 304 2B finish (a mean surface roughness of 0.210±0.227 µm) there was a biofilm

EPS of 8 µg/cm2 on average and on stainless steel 304 2D finish (a mean surface roughness of

1.198±0.041 µm) there was a biofilm EPS of 4 µg/cm2 on average. Meanwhile, when comparing the cell counts on the surfaces, they did observe a higher viable and total cell counts under higher flow rate; for example, after 2 months of exposure, on stainless steel 304 2B finish there was a viable count of 3500 CFU/cm2 on average and on stainless steel 304 2D finish there was a viable count of 5800 CFU/cm2 on average. Considering the microorganism they used, Acinetobacter,

Pseudomonas, Methylobacterium, Corynebacterium and Arthrobacter spp, they suspected that rod-shaped bacteria attached to the surfaces firmer due to the production of EPS at higher flow rate, but there was no direct validation study. Another study explored the flow effect from a fluid mechanics approach (Kraigsley et al., 2014). By using transparent PVC tubing as the flow channel and E. coli as the biofilm formation microorganism, they tested the biofilm formation under different flow rate for both the upstream side and downstream side of the tubing. Their result showed that an optimal flow rate existed to maximize both the upstream progress of the biofilm and the density of the downstream growth. Under this flow rate, E. coli is able to withstand the corresponding shear forces without dispersing. The transparent tubing helped to visualize the morphology of the biofilm, and they observed that the growth of biofilm changed

15 into a more non-uniform structure under higher shear rates, and the attachment seemed to be in patches on the surface rather than the traditional coating the contacting surface contiguously.

They concluded that their biofilm formation model described an along-the-surface growth type model, and not recruiting planktonic cells from the flowing media model, and this further confirmed the notion of categorizing biofilm modeling into reaction-diffusion system modeling.

However, some early literature stated that under both laminar and turbulent flow, the attachment of microorganisms is enhanced (Rijnaarts et al., 1993). More discussions regarding to the flow characteristics in dairy processing equipment are present in the section 3.2 when discussing about the cleaning-in-place of milking system.

2.1.3.4 pH

Several studies focused on the effect of environmental pH on the formation of biofilm on surfaces. One study explored various adhesion forces of three microorganisms onto stainless steel

316 surfaces subjecting them to different solution properties (Sheng et al., 2008). They used two anaerobic strains (Desulfovibrio desulfuricans and Desulfovibrio singaporenus) and one aerobic strain (Pseudomonas spp.) as the biofilm forming microorganism, and the bacterial-metal adhesion force was quantified using atomic force microscopy (AFM) with a cell probe. Results showed that the adhesion of microorganism to contacting surface reached its peak when the surrounding solution pH was near the isoelectric point of the microorganism, namely at the zero net charge. Results showed that at pH 9, the adhesion forces were higher compared with pH 7; mainly because at lower pH, Fe2+ and Fe+3 ions and negatively charged carboxylate groups are attracted to each other more firmly (Sheng et al., 2008). Another study confirmed the previously stated strain-specific characteristics (Mafu et al., 2011). By using different surfaces (hydrophobic and hydrophilic) under different pHs, they demonstrated that there was no significant difference in the attachment for Salmonella Enteritidis, under various substrate types and culture pHs.

However, for Aeromonas hydrophila, E. coli O157:H7, and Staphylococcus aureus, they observed the pH influence on the adhesion forces. From what they presented, it illustrates that by increasing pH from 6 to 8, the adhesion for some microorganisms increased and then dropped (A. hydrophila and E. coli O157: H7) while for some other bacteria (S. aureus), the adhesion kept dropping, possibly due to the same reason of isoelectric point difference.

2.1.3.5 Presence of nutrients

With the presence of nutrients, some study had shown that the bacteria-metal adhesion was reduced due to the formation of organic film which reduced the metal surface wettability

(Sheng et al., 2008). One study used stainless steel preconditioned with the water-soluble aqueous cod muscle extract. Results showed that the preconditioning significantly reduced the attachment of Pseudomonas fluorescens AH2, compared with other bacteria. In addition, when testing other food related microorganisms (S. Enteritidis, L. monocytogenes N53-1, E. coli MG 1655, etc.), they observed similar results in which animal extracts reduced the adhesion of these microorganisms (Bernbom et al., 2009). As to milk and milk components, researcher observed that the presence of lactose and non-casein protein induced the largest number of bacterial attachment among trials varying different contact surfaces and different types of bacteria and milk components; while the presence of whole milk, fat and casein did not significantly increase the bacterial attachment, compared with quarter-strength Ringer’s solution (Speers and Gilmour,

1985). In another study, skim milk was used as the nutrient source (Barnes et al., 1999). They tested stainless steel coupons treated with skim milk and then followed by bacterial suspensions.

The level of bacterial attachment was determined by counting the number of attached bacteria, which included S. aureus, Pseudomonas fragi, E. coli, L. monocytogenes, and Serratia marcescens. The attachment of S. aureus, L. monocytogenes, and S. marcescens was reduced by skim milk, while the other two microorganism (P. fragi and E. coli) attachments were minimal.

When using single component in milk, results were similar: the suspensions of α-casein, β-casein,

κ-casein, and α-lactalbumin were able to reduce the adhesion of S. aureus and L. monocytogenes as well. Exploration into the proportion of nitrogen in the adsorbed films showed that the attached bacteria were inversely related to the nitrogen amount, which demonstrated that the absorbed protein inhibited the initial microorganism attachment.

2.1.4 Control strategies of biofilm in dairy processing equipment

The biofilm in dairy and dairy processing equipment is undesirable, and there have been extensive studies focusing on the potential intervention and the control of biofilm. Based on the different formation mechanism of the biofilm, several approaches have been proposed.

2.1.4.1 Prevention and disruption of biofilm and biofouling

Instead of treating the post contaminated biofilm using certain cleaning and sanitizing methods, it would be more effective to prevent the formation from the beginning. To prevent the microorganism from attaching and biofilm formation, several studies have been conducted. Initial biofilm formation depends on the contact surface properties; the simplest method is to minimize the surface area for potential biofilm formation and shorten the transit time of fluid through the problematic sections.

The clustering of microorganisms onto the contacting surfaces is the beginning stages of biofilm formation. During this rapid attachment, electrostatic forces between the bacteria and the substrate (referred to as “bioelectric effect”) are generally repulsive; this provides a possible opportunity to disrupt the process by applying a small current/voltage and provoking the bacterial detachment. There have been studies showing the inhibiting effect of certain current and voltage against Streptococcus thermophilus attachment (Flint et al., 2000), the desorption of

18 microorganisms from a human salivary conditioned film ( Caubet et al., 2004; Poortinga et al.,

2001; Shirtliff et al., 2005). A recent study tested the in vitro bioelectric effect on 11 antimicrobial agent activities against several representative microorganisms, and results showed no significant biofilm reduction for most of the agent treatments. For some cases using combination treatments, the bioelectric effects were statistically significant. The researchers therefore concluded that the application of electric current as the enhancement of antimicrobial agent activity might not be a generalizable phenomenon across all microorganisms and antimicrobial agents (del Pozo et al.,

2009).

2.1.4.2 Removal of biofilm with addition of surfactant or enzymes

To improve the cleaning and sanitizing performance for biofilm removal, in addition to enhancing the mechanical effect of solution flow, another aspect is to improve the solution physiochemical properties. Researchers demonstrated that the wettability of stainless steel increased when its surface is applied with N-Acetyl-l-cysteine, and this resulted in a substantial reduction in bacterial adhesion and EPS production, thus preventing the biofilm formation

(Olofsson et al., 2003).

One study used commercial α-amylase from different sources and evaluated their ability in reducing and preventing S. aureus biofilm formation (Craigen et al., 2011). They confirmed the capability of α-amylase in inhibiting S. aureus biofilm formation and getting it detached at a rapid rate; in addition, α-amylase reduced and dissociated the bacterial cell aggregation in liquid suspension. The hydrolyzing capability of the enzyme on the glycosidic linkages of EPS contributed to the inhibition effect of cell-to-cell associations. They also speculated that the amylase degraded the EPS precursors which prevented biofilm formation. Another study explored the cleaning effect of enzyme based cleaning agents in removal and inactivation of

Bacillus spp. biofilm (Parkar et al., 2004). Results showed that the enzyme preparations were not 19 completely effective, probably due to the low wettability of these cleaning agents. In addition, they pointed out the ineffectiveness also came from the characteristic of enzyme itself – these enzyme cleaners only target one part of the biofilm, instead of a whole mixture. Addition of surfactants or combination with other agents might achieve a better cleaning efficiency.

2.1.4.3 Removal of biofilm with modified contact surfaces

In addition to modifying the solution physiochemical properties, there have also been studies focusing on the modification of the contacting surface to reduce the microorganism adherence and biofouling. In dairy and other industries, the most commonly used material is stainless steel due to its high heat transfer coefficient, the ease of cleaning and sanitizing, and resistance to corrosive chemicals.

Extensive studies were done to modify the stainless steel surfaces by implanting surface ions and reduce especially the biofilm formation and milk fouling. A graded electroless Ni-P-

PTFE (polytetrafluoroethylene) antibacterial coating was evaluated on the ability to reduce the bacterial adhesion on metallic surfaces (e.g. stainless steel) (Zhao, 2004). The coating was able to reduce the attachment by 82-97%, using an intermolecular force theory. Another study demonstrated that when coated with Ag-PTFE, stainless steel surfaces were able to reduce the attachment of E. coli by about 2 logs compared with silver coating, stainless steel alone or titanium alone (Zhao et al., 2005a). In addition, they proved that the existence of silver enhanced the corrosion resistance (in 0.9% NaCl solution) compared with stainless steel 316L. Another coating the researchers tested was a cost-effective autocatalytic graded Ni–Cu–P–PTFE composite (Zhao et al., 2005b). The coated surfaces had a significant impact on both the bacterial attachment and the mineral fouling (CaSO4 deposit) adhesion when the surface free energy of the coated surfaces was within a certain range. When stainless steel 304 plates was modified with Ni-

P and small amounts of PTFE (Zhao and Liu, 2006), the reduction of E. coli attachment was 20 about 1 log compared with non-coating plates. In addition, the investigation of effects of pH, temperature, and PTFE concentration showed that these other surrounding and environmental factors exerted significant impacts on the bacterial adhesion as well. For heat exchanger systems, extensive studies were conducted as well. Several types of coatings were evaluated compared with stainless steel, such as LectrofluorTM-641, Ni-P-PTFE (Balasubramanian and Puri, 2008b;

Balasubramanian and Puri, 2009a), and AMC-148-18 (Balasubramanian and Puri, 2008a;

Balasubramanian and Puri, 2009b). By changing the flow rate of the testing fluid (tomato juice or milk), the foulant mass was measured and compared. All the coated plates exhibited a reduced fouling amount resulting in more energy savings and lower operation cost. In addition, the morphology of foulants on coated surface was seen to be loosely held and more amorphous compared with the ones on stainless steel (more crystalline) (Balasubramanian and Puri, 2008b).

2.1.4.4 Cleaning methods

The most time-saving and important method in controlling the formation and adherence of biofilm during dairy practices is to clean the surfaces. Traditionally, the cleaning was completed by hand using a brush; it was therefore categorized as cleaning-out-of-place (COP), commonly used by small scale industries. However, it is less efficient and laborious process. On the other hand, if the pipelines need not to be disassembled during the cleaning, the process is referred to as cleaning-in-place (CIP). The advantage of CIP is that the cleaning and sanitizing process takes place inside the pipelines, which is highly automated resulting substantial time savings over COP. More information of the CIP process for especially the milking system is explained in detail and discussed in section 3.2.

For the COP method, instead of hand cleaning and sanitizing everything, more recent studies focused on ultrasonic cleaning and its efficacy when combined with other technologies.

Ultrasound is a form of energy generated by sound waves which is transmitted as pressure waves 21 of frequencies that are too high to be detected by human ear, (above 16 kHz) (Jayasooriya et al.,

2004). The ultrasonic cleaning to remove biofilm or foulant happens when cavitations occurs due to pressure change in the fluid medium, or sometimes from the chemical interactions with ultrasonic generated radicals (Kallioinen and Mänttäri, 2011). One research group used E. coli and S. aureus as the contaminant for milk, and tested the biofilm removal effect of ultrasound and its combining effect with other chelating agents (Oulahal et al., 2004). Two different ultrasonic devices were used, one transducer was flat and one was curved. Results showed that when biofilm was developed on opened stainless steel surfaces, the flat transducers were able to remove the biofilm formed on stainless steel surfaces effectively at 10 s of treatment and 40 kHz.

However, when a closed surface was tested using the curved transducer, the biofilm removal was not satisfactory; 30 and 66% for E. coli and S. aureus biofilms, respectively. To improve the removal performance, the combining effect was applied with chelating agent. Complete removal was observed for E. coli but not for S. aureus. This again confirmed the previously mentioned characteristic of biofilm formation, which means the formation mechanism, is species and strain specific. The synergistic effect of ultrasound and ozone was also investigated (Baumann et al.,

2009). They developed the biofilm using L. monocytogenes and studied its removal using

“power” ultrasound (20 kHz, 100% amplitude, 120W) alone and when combined with ozone.

There was a 3.8 log CFU/ml reduction when treating the inoculated stainless steel chips for 60 s, but when combined with a following ozone treatment at 0.5 ppm ozone, a 7.31 log CFU/ml reduction was achieved. Obviously the “combo” effect is much better than single treatment, and more research should focus on the combined treatments for biofilm removal with ultrasonic cleaning. The limitation of ultrasonic treatment lies in the scale of treatment, because target parts have to be disassembled for cleaning. This is not only inconvenient in large processing plants, but also costly due to the longer downtime. The development of in situ application of ultrasonic would be an interesting topic for future research.

2.1.5 Monitoring and detection of biofilm in dairy processing equipment

Despite the advanced technology developed for controlling and removing biofilms, the monitoring and detecting of biofilm formation is essential to implement the control and removal technologies and prevent any further biofilm formation.

The most traditional and direct methods besides visual appearance by naked eye is the bacterial enumeration. There are various sampling and enumeration methods for biofilms in dairy plants such as swabbing, rinsing, agar flooding, and agar contact (Kumar and Anand, 1998).

However, there are limitations to these agar plate count methods, and the most relevant is the requirement for 18 to 48 h incubations and this time delay is not practical. Furthermore, the microorganisms found in the biofilm in a dairy environment are subjected to various stresses such as starvation, chemical interactions, sudden temperature increase or decrease, mechanical shear forces (Marchand et al., 2012), which causes surviving microorganisms to be unculturable, and thus these agar plate count methods underestimates the actual population inside the biofilm

(Wirtanen, 1995).

There are other methods to monitor the growth of biofilm and biofouling. One study used two turbidity measurements to monitor the fouling inside the heat exchanger surfaces, and a commercial sensor was developed (Klahre and Flemming, 2000). Another sensor measures the heat flux and the temperature on the hot side of the heat exchanger. The fouling rate is then calculated from the heat transfer coefficient normalized to its value at the beginning of the run

(Marchand et al., 2012). A fouling cell assembly in stainless steel was developed as a replacement in dairy pipelines (Fornalik, 2008). It is able to monitor the biofilm accumulation without removing the equipment and it is used as an objective indication of the on-farm CIP effectiveness.

Another study used an ultrasound based sensor system to determine the presence, or absence of fouling within a planar heat exchanger (Wallhäußer et al., 2013). The ultrasonic signal was then analyzed and fed together with temperature and mass flow rate measurements for decision of

23 fouling presence. High accuracies were achieved for both classifications – artificial neural network (ANN) with more than 80% and support vector machine (SVM) with more than 94%.

However, these models were not tested for any industrial applications.

To achieve in-line detection, a mechatronic surface sensor was developed (Pereira et al.,

2008). It utilized the propagation of nano-vibrations along a surface to monitor the deposit establishment. By comparing the fouling curves of three deposits including the P. fluorescens biofilm formed under turbulent flow and the P. fluorescens biofilm formed under laminar flow, along with the silica deposit formed under turbulent flow from the sensor, a good fit was observed between the sensor output signal and the deposit build-up on the surfaces, with silica deposit slightly below the other two. Another comparison between laminar and turbulent flow with the same microorganisms showed a higher deviation under laminar flow, and they suspected the variation might come from the instable biofilm physical structure under lower shear stresses. A highlight for this research was the sensor’s capability to distinguish biological deposits from inorganic ones – this is of considerable importance to most of the food processing procedures

(especially dairy processing), the bacterial biofilm and fouling are always a combined effect resulting in the deposit on the surfaces, while in fact the original sources for the deposit are different. They further investigated the capability of the sensor when applying it to a tubular configuration. Simulated milk deposits were used with calcium phosphate and protein fouling agents (Teixeira et al., 2014). Results showed the sensor was capable to distinguish the natural differences of the deposits besides indicating the deposit buildup. The parameters they set up clearly showed distinctions between a more rigid deposit (i.e., simulated milk ultrafiltrate) and more viscoelastic ones (i.e., whey protein isolates or a mixture of simulated milk ultrafiltrate and whey protein isolates).

Quick off-line detection is equally or more important since the initial expensive investments needed for the in-line detection methods. Adenosine triphosphate (ATP)

24 bioluminescence method is one of the most widely studied methods in recent years. ATP bioluminescence is a rapid cleaning testing method, and it has the advantages of fast detecting and easy handling. ATP can be found in all viable cells from plants, animals, and microbial (Leon and Albrecht, 2007). The level of ATP can be measured by the reaction of luciferin and luciferase enzyme complex which results in generating light. This light can then be detected and measured by a luminometer in an expression of Relative Light Unit (RLU) readings (Chen and Godwin,

2006). Given the fact that all living cells are capable of generating ATP, RLU readings can be utilized as an indicator of the biological residual which indirectly reflects the cleanliness. ATP bioluminescence technology is now widely used as a real time estimate of cleaning for the food processing equipment and food contact surfaces (Ogden, 1993; Larson et al., 2003).

There are several studies using RLU readings to indicate milk quality and the milking system cleanliness. Vilar et al. (2008) further improved this method of using ATP bioluminescence for the milking equipment surface cleanliness evaluation. They examined the surfaces of teat cup rubbers, teat dip containers, milk receivers, and pipeline joints using both bacterial counts and RLU readings. One highlight of their work was that they established the cut- off points for different sampling locations separately, based on the scatterplot of log10BC

(Bacterial Counts) and log RLU. These reference values differed from log 2.28 for teat cup rubbers to log 3.26 for pipeline joints; but they share similar percentages of around 65% among their own sampling groups. Another highlight from their study was that they established a multiple linear regression model for the prediction of bacterial counts of the bulk tank from the independent variables of RLU readings from teat cup rubber, teat cup container, milk receiver and pipeline joint with a R2of 0.12. Despite the unsatisfactory correlation coefficient of the regression equation, this study provided the basic idea for using independent variables from the milking system sampling to predict the cleanliness of other surfaces within the milking system which were difficult to draw samples from for direct evaluation. In addition, this attempt brought an idea

25 of correlating the indirect RLU readings with other indicators to build a more detailed and comprehensive description for the evaluation of cleanliness of the milking system.

Cais and Pikul (2008) conducted similar experiments using the bioluminescence method.

They compared two types of plate surfaces used in a dairy with different surface roughness (Ra) and built a linear regression model of RLU readings of microbial counts for each surface. The correlation coefficient was high; one was 0.975 (with the surface mean roughness Ra=0.6) while the other was 0.987 (with the surface mean roughness Ra=0.8). In addition, they also developed a classification of clean, conditionally clean, and unacceptable based on different RLU reading ranges. A recent study conducted by Carrascosa et al. (2012) compared the results of using dip slide method, contact agar plate method and ATP measurement in the dairy production chain

(curd vat, filler, mould, table, and tank at five different cheese factories). The terms of frequency and percentage were introduced to describe the relationship among these methods, but no regression model was provided among these methods.

Despite the improvement made these years on using ATP bioluminescence to evaluate surface cleanliness, the limitation of this technology should be paid attention, too. Turner et al.

(2010) conducted a series of experiments to determine the limit of ATP detection for different microbial and organic contaminants. By serial dilutions of Escherichia coli, Staphylococcus aureus, Toxocara canis, Toxoplasma gondii tachyzoites, epithelial cells, and rodent blood, urine, and feces, the ATP detection system was examined. Results showed that for the ATP limit differed based on various contaminants; the lower detecting limit ranging from 102 CFU/10µl of S. aureus

(representative gram-positive bacteria) to 104 CFU/10µl of E. coli (representative gram-negative bacteria). In addition, the study also brought up a noteworthy phenomenon, which is the additional physical effect on the sensitivity of this ATP testing system. Results showed that the efficacy of this ATP-based system was adversely affected in the presence of residual disinfectants;

26 the application of sonication of contaminants which led to a more complete cell lysis would enhance the detection sensitivity.

2.2 Cleaning-in-place for milking system

The most common and effective practice on a commercial dairy processing plant to prevent biofilm and biofouling formation is to clean and sanitize the contact surfaces on a regular basis; this would reduce the chance of bacterial attachment and deposition (Simões et al., 2006).

The cleaning and sanitizing program should be effective in a way that microorganisms and undesirable residues are removed from the surfaces to the utmost while keeping the cleaning and sanitizing solution usage to the lowest from an economic aspect (Dosti et al., 2005). Despite the development of COP approaches mentioned above, more focus has been paid to the cleaning-in- place method, because of its high automation and potential scalability in larger food processing plants. CIP systems have the advantages of energy and cost saving; less labor requirements, which can be easily transformed into large-scale industrial production; longer operation, longevity due to less disassembly frequency; and reduction in the hazardous operation process dealing with the handling and inhaling of cleaning chemicals (Bowser et al., 2005).

CIP is applied in most milk handling and processing lines. In the United States, there are a series of recommendations from Dairy Practice Council (DPC) regarding to the detailed CIP procedures for different types of milking systems. In general, the basic CIP consists of the following steps: i) warm water rinse right after the milking is finished to remove the gross residues; ii) hot chlorinated alkaline wash to remove the organic matter residing on the milk line surfaces; iii) intermediate water rinse to flush out the detergent and remove minerals; iv) acid rinse (usually unheated) to inactivate and kill any remaining microorganisms; and v) a sanitizing rinse right before next milk. Details of these steps are provided below based on recommendations from Dairy Practice Council (DPC, 2010): 27

i. The warm water rinse begins right after the milking process is finished. This step can

remove the residual milk remaining in the system. The temperature of the pre-rinse water

should be confined between 46.1 to 51.6°C, and the reason for the highest limit (51.6°C)

is to prevent protein denaturing, which might result in the protein clinging to the system;

the reason for the lowest temperature limit (46.1oC) is to be above the melting point of fat

to ensure that fats will be removed and not re-deposited. These rinses should not be re-

circulated. It is recommended that after every chemical cycle there should be a water

rinse to remove the organic and chemical residuals in the system and the equipment. ii. The next step is alkaline solution wash. Typically, the alkaline wash solution consists of

sodium hydroxide (caustic soda) potassium hydroxide (caustic potash), sodium carbonate

(soda ash), and sodium silicates. Trisodium phosphate (TSP) is also placed into the alkali

group because of its reaction with water to yield hydroxide ions (DPC, 2006). In addition,

various sources of available chlorine are very often added to alkaline cleaners. These

enable the cleaner to solubilize protein residues by peptizing the protein and to promote

rinsing of the equipment, leaving it free of water spots (DPC, 2006). Due to the high

chlorine content, this cycle functions both to remove organic soils such as milk fat and

proteins and possesses the potential of killing bacteria. The pH of the alkaline solution

should be around 12 with a chlorine content of 120 ppm and an alkalinity of 1100 ppm.

Generally, the temperature of this alkaline wash is between 71.1 to 76.7°C, and the

cleaning effectiveness depends on the relatively higher temperature and lower supply

water hardness. In addition to that, it is required that the drain temperature of this alkaline

solution water should be higher than 48.9°C to ensure good cleaning efficiency. As

pulsing air and fluids which create fast moving turbulent flows (slugs) would clean the

milking systems during rinsing, washing and sanitizing cycles, it is recommended that the

alkaline solution should be circulated around the system for about 8 to 10 minutes with a

minimum of 20 slugs. iii. Between the alkaline wash and acid wash cycles, a tepid water rinse cycle is preferred. It

is only one pass then drained, to remove the detergent from the alkaline wash and prevent

potential mixing of remaining chlorine and the acid wash chemicals. iv. After the alkaline wash, the acidic rinse, with a pH around 3 and solution usually

unheated, serves to remove the mineral deposits from water and milk, neutralize the

alkaline cleaner, and leave the system in an acidified state to retard bacteria growth

(Reinemann & Ruegg, 2000). Phosphoric acid is the most commonly used inorganic acid

because it is effective, relatively safe, and less corrosive than other mineral (inorganic)

acids. Hydrochloric acid, nitric acid are also used accompanied with phosphoric acid

(DPC, 2006). It is important that the acidic rinse solution drains completely from the

system after the completion of the acidic rinse cycle. - v. The last sanitizing cycle, usually circulated immediately before milking, is to kill bacteria

residing in the milking system (Reinemann et al., 1995); this last sanitizing cycle is also

required to be drained completely (DPC, 2006). . Hypochlorites, elemental chlorine,

iodophors, mixed halogens are often used for sanitizing purpose in fluid milk processing.

The most common type of chlorine sanitizers used in the dairy industry is hypochlorite.

Figure 2-2. Tie-stall barn milking system at a commercial dairy farm. (a). Barn view with milk

pipelines above the cows around the barn; the 3D Solidworks® drawing shows a typical

commercial dairy farm milking system pipelines around the barn and in the milk house.

(b). Teatcups and claws in the solution sink in the milk house in a typical commercial

dairy farm.

For small scale commercial dairy farms, most pipeline milking systems share similar design features. The interface to the individual animal is the milking unit. The milking unit consists of a cluster of four teat cups and a claw (ASABE Standards, 2008). Each teatcup is made up of a rigid shell with a rubber liner inside (fig. 2-2b). The rubber liners are connected to the claw, which is simply a manifold that connects the four teatcups to the pipeline. The pipeline, or milkline, carries the milk from the milking unit to the receiver. The pipeline is typically stainless steel and can range from about 3.80 to 10.16 mm in diameter depending on the size and capacity of the milking system. As is shown in figure 2-2a, the pipelines, above the cows’ heads, extends 30 around the barn and the cows are milked at their stalls. The Solidworks ® drawing in figure 2-2 presents the pipeline configuration at this particular dairy farm in the barn and in the milk house.

In such a tie-stall system, the milking units are moved from animal to animal and connected to the pipeline via an opening in the pipeline called a milk inlet or nipple. The receiver, typically glass or stainless steel, is a vessel that receives milk from one or more pipelines and connects to both the vacuum supply and the milk pump. The milk pump, or releaser milk pump, removes milk from the receiver under vacuum and discharges it at atmospheric pressure (Walker et al., 2005b).

Most recently, there is another CIP process called “one-step CIP cleaning” and it has received attention on increasing numbers of dairy farms (Parr, 2013). This “one-step CIP cleaning” is a CIP process in which the previous alkaline wash cycle and acid rinse cycle are combined together into one cycle, functioning clean the organic deposits like fat and protein and at the same time remove the inorganic minerals clinging to the surface. Several commercially available one-step CIP chemicals are on the market and their CIP performance is claimed to be comparable to the conventional separate alkaline wash and acid wash of the milking system. It was claimed that using one-step CIP can be cost and time saving through eliminating the separate alkaline and acid wash cycles, decreasing the usage of water, chemicals, energy, and time. In addition, the manufactures claimed that their products are capable of keeping the bacterial count low for standard tests. However, there are no published experimental results for the milking system CIP using these chemicals.

Regardless whether CIP process is a two-step or one-step CIP, one should always keep close attention to the important parameters affecting the CIP performance. One of them is the introduction of air into the wash solution, or “slugs”. The number of slugs should be calculated accordingly as stated above; in order use less water and develop 10 to 20 times higher shear stress

(Reinemann, 2002). One study used turbulent two phased flow (TTPF) to study its ability in eliminating biofilm from the inside of dental tubing and other long tubular channels (Benjamin

31 and Labib, 2000). The other parameter is the rinse or wash solution temperature. As stated above, it is of high importance to achieve the recommended solution temperature to accomplish the desired CIP performance. However, if high temperature is not always available or maintainable, other approaches may be used as an alternative. One study used temperature step changes to make up the potential incomplete cleaning during CIP (Knight et al., 2004). By establishing a pilot plant for cheese-milk pasteurization, the researchers studied the effect of temperature cycling on the development of S. thermophilus biofilm. A 106 CFU/ml level of S. thermophilus was detected under normal conditions on the cooling sections of the plant; however, by applying a temperature changing steps periodically to the growth region of S. thermophilus, the biofilm development was controlled and positive effects even observed on the heating side of the plant. They came to an optimal step change to 55°C, applied for 10 min, and with a 60 min interval between step changes, which could result in a 20-h production run with no detectable S. thermophilus growth.

Provided an adjustable temperature setting and control system is available, this temperature fluctuation seems promising in controlling other microorganisms and biofilm formation as well.

2.3 Electrolyzed oxidizing (EO) water

The chemicals used during the CIP process are potentially hazardous and environmentally harmful; therefore, more studies had been conducted in recent years to find alternatives for the conventional chemicals. Electrolyzed oxidizing (EO) water is one of these.

Technology for industrial use of the EO water was initially developed in Japan (Huang et al.,

2008). EO water is generated by electrolyzing a diluted salt solution in an electrolytic chamber with a membrane between the anode and cathode in the middle (fig. 2-3). The direct current drives negatively charged ions to the anode ending in oxygen, chlorine, hypochlorite ion, hypochlorous acid and hydrochloric acid; while positively charged ions go to the cathode resulting in hydrogen and sodium hydroxide (Hsu, 2005). The anode will generate acidic EO

32 water with the cathode generating alkaline EO water at the same time. The acidic EO water has the potential to reach a pH as low as 2.6 and an oxidation-reduction potential (ORP) as high as

1,150 mV. The alkaline EO water possesses a pH as high as 11.4 and a negative ORP to -795 mV.

Depending on needs, different properties of EO water can be generated by adjusting amperage and voltage.

EO water generators are produced by several companies (fig. 2-4and 2-5); despite the different configurations of the EO water generator, the EO water generation mechanism is similar.

The principle of generating electrolyzed oxidizing water is shown in figure 2-5 with the following chemical reactions occurring (Huang et al., 2008):

Figure 2-3. Schematic of electrolyzed water generator and produced compounds at both

electrodes (Huang et al., 2008).

Figure 2-4. EO water generator (Model ROX60SA, Hoshizaki Electric Co. Ltd, Japan).

Figure 2-5. Different configurations of EO water generator. (a). Viking Puri: Mini Puri (Viking

Puri, LLC, FL) (b). Aquaox ECS300 system (Aquaox BV, Netherlands).

2.3.1 Advantages and disadvantages of EO water

Compared to other cleaning agents, EO water has some advantages. The main advantage is its safety. Study by Sakurai et al. (2003) demonstrated that EO water serves as an appropriate approach to clean and disinfect digestive endoscopes between patients, and this method had less adverse impact on human body and the environment, comparing to traditional chemical methods.

In addition, is the researchers showed that the cost of using EO water is much less compared with using glutaraldehyde ($0.05 or less per liter). For milking system CIP usage, once an EO water generator is purchased, the only operating expenses are water and salts to generate the EO water and electricity (Walker et al., 2003).

However, there are also some disadvantages of EO water, and the major one is that study showed a loss in the antimicrobial activity when the EO water, especially the acidic EO water,

35 was stored for seven days or more (Nagamatsu et al., 2002). In addition, due to the extremely low pH and high free chlorine content of acidic EO water, the chlorine gas emission, metal corrosion, and synthetic resin degradation cause some concerns as well (Huang et al., 2008). To date, the cost of an EO water generator machine is still relatively high to be an obstacle for further infusion of this technology, and currently there is no available long term evaluation of using EO water for cleaning and disinfecting purposes in industry.

2.3.2 EO Water used as a disinfectant in the food industry

2.3.2.1 Acidic EO water

Acidic EO water has been demonstrated to have satisfactory sanitizing and disinfection effects. There are several indicators for comparison when evaluating the quality of acidic EO water, including: pH, ORP, electrical conductivity, total residual chlorine, dissolved oxygen, sodium ion and chlorine ion concentrations, etc. (Hsu & Kao, 2004). Buck et al. (2002) showed that germination of all the 22 tested fungal species was significantly reduced or prevented under the treatment of acidic EO water. Okull and Laborde (2004) showed that if wounded (surface peeled off) apples are contaminated, EO water reduced the viable spore population by greater than 4 log units (50% EO water greater than 2 log units), but was not able to prevent lesion formation on fruit previously inoculated with microorganisms. Another study used spray acidic

EO water to reduce the microbial load on hatching eggs, and result showed that by spraying acidic EO water on the eggshell before incubation, the microbial load was significantly decreased; no adverse effect was observed on cuticle structure, normal embryonic development, and hatchability (Fasenko et al., 2009). Similar results showed the microorganism reduction potential against E. coli K12 of using acidic EO water for the egg washing process for a pilot scale study (Bialka et al., 2004). Several studies reported the effect of using acidic EO water treating seafood. Huang et al. (2006) soaked E. coli and Vibrio parahaemolyticus inoculated 36 tilapia samples into acidic EO water, and result showed a reduction of 1.68 and 3.84 log10

CFU/cm2, separately. Ozer and Demirci (2006) showed that by using a treatment process: i)

Alkaline and acidic EO water treatment and ii) Acidic EO water alone treatment, on the E. coli

O157: H7 and L. monocytogenes inoculated salmon fillets, results were comparable with the results of the control group by using 90 ppm chlorine solution treatment.

There are also studies using acidic EO water combined with other technologies to achieve satisfactory disinfecting effects. Okull et al. (2006) showed that three nonionic surfactants, polyoxyethylene sorbitan monooleate (Tween 80), polyoxyethylene sorbitan monolaurate (Tween

20), and sorbitan monolaurate (Span 20) will not improve the EO water performance against spores of Penicillium expansum in aqueous suspension. Buck et al. (2002) showed that triton X-

100 (at all concentrations) and Tween 20 (at concentration of 1% and 10%) would eliminate the activity of EO water against Botrytis cinerea. Huang et al. (2010) used additional ultrasonication as a complement to enhance the inactivation of E. coli O157:H7; inactivation reached to almost 2 logs reduction.

Another aspect that scientists focused on is the EO treatment time duration, meaning the contact time of EO water with products. Fabrizio and Cutter (2004) showed that a 15 s spray with acidic EO water has the ability to reduce Campylobacter coli associated with fresh pork surfaces with a significantly reduced population of 1.81 log CFU/cm2. Serraino et al. (2010) used the acidic EO water within a week after being produced, and they showed the changes in acidic EO water characteristics with time after spraying on steel and Teflon surfaces. However, there is no direct indication that the longer contact time duration with acidic EO water would result in a better cleaning and disinfecting effect. Ren and Su (2006) showed that longer time treatment (>12 h) of oysters using acidic EO water will cause high levels of residual chlorine concentration, which is detrimental to the oysters. This result showed that the duration of the treatment is essential, and residual chlorine might be left over if handled improperly.

2.3.2.2 Near-neutral EO water

The low pH of acidic EO water might have an adverse effect on the processing equipment and could cause a certain amount of surface corrosion if utilized for a longer time period.

Therefore, near-neutral EO water has been studied recently. Near-neutral EO water has a pH of around 7 and still maintains a high ORP of around 700 mV. Guentzel et al. (2008) showed that using near-neutral EO water to clean the surfaces containing mixed heterotrophic bacteria on spinach and lettuce is effective, and this EO water solution can also be used for rinse treatment to reduce the bacterial populations. Similar results can be found from the study of Abadias et al.

(2008), in which diluted near-neutral EO water containing approximately 50 ppm of free chlorine with a pH of 8.6. The population reduction of E. coli O157:H7, Samonella, Listeria innocua and

Erwinia carotovora on the lettuce was similar to using chlorinated water with free chlorine concentration of 120 ppm. This study indicates that near-neutral EO water is a promising alternative disinfection agent in food industry when it comes to disinfecting fresh-cut produce – using EO water can achieve similar microbial reduction effect as sodium hypochlorite obtained while reduce the usage of free chlorine at the same time. In another study, different concentrations of total residual chlorine from 25-100 ppm of the near-neutral EO water were used to treat organisms of Botrytis cinerea and Monilinia fructicola (Buck et al., 2002). Results show that this near-neutral EO water solution may be effective for the postharvest sanitation of fruit surfaces.

There is an additional result demonstrating the advantage of using near-neutral EO water instead of acidic EO water: strong acidic EO water has antifungal properties, but the chlorine will be lost at a faster speed than the near-neutral EO solutions containing primarily HOCl, which will not lose chlorine rapidly (Guentzel et al., 2010). These results offered a better method of disinfection and sanitization. Another study conducted by Guentzel et al. (2011) revealed the possible bacterial-killing principle: the high oxidation reduction potential (ORP) of near-neutral EO water affects the outer membrane of the fungi, leading to further oxidation of intracellular reactions and

38 respiratory pathways. Furthermore, they explained two possible functions of near-neutral EO water: manage infection of B. cinerea on strawberry plants in the field and a disinfection solution for facilities related to this process to prevent or manage infections of B. cinerea and M.

Fructicola. Due to the low chlorine content of the near-neutral EO water, it is proved that no significant residue remained (p>0.05) when comparing to other bleach or high chlorine EO water treatments. In addition, a hypothesis that the difference between near-neutral EO water and acidic

EO water originates from the OH radical was verified through UV spectroscopy, 17O-NMR spectroscopy and electron spin resonance analysis. More OH radicals exist in the near-neutral EO water than that is in the acidic EO water. Conclusion was drawn that different OH radical levels is the key to fungicidal efficiencies against Aspergillus flavus (Xiong et al., 2010). Audenaert et al.

(2012) showed that when using near-neutral EO water treating Fusarium graminearum spores, there is an increasing deoxynivalenol biosynthesis level on the sub-lethal amendments while reducing F. graminearum efficiently in vitro which indicates that near-neutral EO water has the potential to control Fusarium spp. in wheat grains during transportation and storage. More studies should be focused on the material surfaces cleaning and disinfecting using near-neutral EO water; these materials actually have direct contact with the food produce during food processing procedures and the cleanliness of their surfaces matter, too.

2.3.2.3 Other combinations

Several different combinations of acidic and alkaline EO water were studied by

Stevenson et al. (2004), and results showed that the alkaline EO water did not inactivate E. coli

O157:H7 at concentrations up to 16% (anode: cathode, vol:vol). Bialka et al. (2004) used the method of soaking the eggs in the alkaline EO water followed by soaking them in the acidic EO water. They also found that when treated alone using either alkaline or acidic EO water, the acidic

EO water performed better in killing bacteria than the alkaline EO water (Bialka et al., 2004). 39

Similar results were proved by Fabrizio et al. (2005). They found that acidic EO water was effective in inactivating L. monocytogenes by dipping at 25°C for 15 minutes, and the combination of acidic and alkaline EO water resulted in a slight reduction of L. monocytogenes by spraying on frankfurters and ham. Likewise, more studies should focus on the cleaning and disinfecting effect of these combinations on the hard surfaces of food contact materials

2.4 Using EO water as an alternative CIP approach for milking system

The cleaning and disinfection efficacy of EO water had been demonstrated and improved by researchers; as a result, using EO water as an alternative method for the milking system CIP is gathering attention in recent years. Walker et al. (2005b) conducted preliminary trials to assess the feasibility of using EO water for CIP of milking systems. Initially they evaluated the efficacy of acidic EO water in cleaning for five milking system-related materials; stainless steel sanitary pipe, PVC milk hose, rubber liners, rubber gasket material and polysulfone plastic. They developed a response surface model to predict the optimal treatment time within the range of 25 to 60°C, for the effective use of EO water for the CIP of milking systems. The temperatures investigated by them were lower than the temperatures currently used (around 70 to 75°C,) for the conventional dairy cleanser (DPC, 2010), which could prove the CIP using EO water to be a less energy-consuming process. However, the exact energy efficiency and cost effectiveness still need to be quantified to provide more convincing evidence. Walker et al. (2005b) evaluated both the short term (single cycle of milking and cleaning) and long term (10 consecutive cycles of milking and cleaning) EO water treatment results using a pilot-scale milking system. Results showed no significant differences in cleanliness between EO water cleaning (7.5 and 10 min) and the conventional cleaning. However, they did not reach a definite conclusion about the time duration needed when carrying out the long-tern study and the effects. A pilot-scale milking system with all major components of a typical pipeline milking system is shown in figure 2-6.

Figure 2-6. Schematic of pilot-scale milking system used by Walker et al. (2005b).

Despite being conclusive of the effectiveness of the EO water CIP of the milking systems, Walker et al. (2003) did not take into account the fact that the acidic EO water has no need to be heated to a temperature as high as the alkaline water. Such high heating temperature will result in significant loss of chlorine and thereby decreasing its effectiveness as a sanitizer.

Moreover, the cleaning effectiveness of the alkaline EO water could be improved at temperatures higher than the temperature of 60°C, which was investigated by Dev et al. (2014).

Temperature of the alkaline and acidic EO water was optimized experimentally with the aid of statistical (response surface) modeling followed by mathematical modeling using the conventional heat transfer calculation. A logarithmic mean temperature of 58.8°C and 37.9°C for the alkaline and acidic EO water, respectively, was established as the optimal temperature for effective CIP of the milking system. An approximate optimal starting temperature for the alkaline and acidic EO water can be determined based on the total surface area of the stainless steel

41 pipelines and the ambient temperature of the immediate surroundings of the milking system by using the conventional heat transfer calculations. Thus, the CIP control system can be programmed to perform CIP effectively using the EO water based on the pipeline configuration of the milking system thus assuring maximum cleanliness under all circumstances (Dev et al., 2014).

However, despite the research on near-neutral EO water for the disinfection on food product surfaces, there is no research conducted on using near-neutral EO water solution to clean the milking system, either in a pilot scale system or on a commercial dairy farm. Given the properties of this chemical and the published studies of near-neutral EO water and the rising trend of commercial “one-step” CIP, using near-neutral EO water for “one-step” CIP for milking system is worth exploration.

Despite the development of using EO water for the milking system CIP, there are still more questions to be answered to further improve and promote this technology, making it eligible for the market in the end. Specifically, i) long term real world practice is necessary to evaluate the stability of using near-neutral EO water for milking system CIP; ii) use of near-neutral EO water for one-step cleaning effectiveness of milking system needs to be investigated with reference to the key EO water and process parameters. Toward that end, optimization of using near-neutral EO water for one-step milking system CIP is needed; iii) mechanisms behind the cleaning process - detailed contribution of each cleaning cycle to the whole CIP cleaning process should be studied and quantified using a mathematical model.

2.5 Milk fouling and CIP process models

Another aspect where the CIP process is often applied to is the removal the fouling deposit on the milk processing equipment surfaces. Fouling happens in food industry when thermal treatments are applied leading to the changes of physiochemical properties of the products and substances deposition. The components of deposition vary in a large range; both 42 organic and mineral could be deposited (Wilson et al., 2002). Milk fouling is a great concern for the milk processing equipment.

2.5.1 Milking fouling mechanisms

A current recognized fouling model is shown in the figure 2-7, which was adapted from the previous work done by Toyoda and Fryer (1997). In this model, proteins react in both the bulk and the thermal boundary layer in the milk. Native protein N is transformed to denatured protein

D in a first order reaction. The denatured protein then reacts ending aggregated protein A in a second order reaction. N*, D*, and A* are the native, denatured, and aggregated protein in thermal boundary layer, respectively. Aggregated protein in the thermal boundary layer, A*, adheres to surfaces causing fouling deposits (Jun and Puri, 2004).

In addition to the protein fouling, there were studies on the calcium phosphate fouling as well. Calcium phosphate deposition and adhesion on a solid surface is a complicated process. It starts with calcium phosphate particle formation when subjected to thermal treatment. Depending on the established forces between foulants and surface, these calcium phosphate particles would adhere to the surface to form the first layer. Then other particles would adhere to the top of this layer and began the growth of the second layer of deposits (Rosmaninho et al., 2007).

Figure 2-7. Description of hypothesized fouling model involving the interactions of different

stages of proteins in the bulk fluid, in the thermal boundary layer and on the contact

surface (N: native protein; D: denatured protein; A: aggregated protein; superscript *

represents a second order reaction and no surperscript represents a first order reaction)

(Georgiadis and Machietto, 2000).

Despite the development of a range of possible surfaces that can affect deposition and furthermore reduce the level of fouling (Rosmaninho and Melo, 2006; Ozden and Puri, 2010), more work should focus on the development for the detachment and removal of food product deposition on surfaces. As stated above, the whey protein constitutes foulants in pasteurization equipment and several other dairy processes (Burton, 1968). In the study of cleaning these milk protein deposits, a three stage process was proposed by Grasshoff (1999). As is shown in the figure 2-8, these three stages are, (a). a swelling stage, during which the native protein reacts to form a high voidage matrix structure; (b). a ‘uniform’ stage, during which the removal of swollen deposits is constant and (c). a decay stage, during which the protein matrix breaks down and patchy deposits are removed from the surface. Understanding the mechanism of these stages are

44 important in better improve the cleaning process to achieve a better cleaning effectiveness, and also help to establish computational fluid dynamics model to simulate the cleaning process.

Figure 2-8. Schematic of the stages involved in removal of whey protein deposits: (a) swelling;

(b) uniform; (c) decay phases (Gillham et al., 1999).

2.5.2 Deposit removal models

Chemical usage will surely affect the cleaning effects in terms of deposition removal.

Christian and Fryer (2006) showed that for the typical cleaning process of fouling plate heat exchanger, no removal would happen unless the deposit was pretreated with chemicals in the first place. In other words, only after chemical treatment, it was possible to remove some of the deposit by other methods, either continue to use chemical treatment or use water rinse. They reached the conclusion that this rinsing cycle after chemical treatment did not bring the system

45 back to total clean again for most conditions, so another pulse of chemical cleaning was needed to remove the final layers of deposits. In addition, they established a simple model to estimate the fraction deposit thickness as a function of deposit exposure time:

푥 (푈푐⁄푈푝) − 1 (∅푐⁄∅푝) − 1 = = (1) 푥푓 (푈푐⁄푈푓) − 1 ∅푐 − 1

In equation 1, where, x was the thickness of the unswollen deposit left on the surface, xf was the initial deposit thickness, U represents heat transfer coefficient, while Uc if at the condition of fully cleaned, Up was partial cleaned, Uf was the fouled condition. Preliminary data suggested that the decrease of thickness with time was linear at the NaOH concentration of 0.5 and 1.0%, meaning that the rate of removal of the deposit was approximately constant. However, at NaOH concentration of 0.1%, the rate of removal did not show strong linear relation, which indicated that certain chemical concentration was of necessity for thorough cleaning actions.

For milk heat exchangers, the deposit cleaning process was firstly studied by Dürr and

Graßhoff (1999) and they proposed a two-factor exponential model. The remaining soil,

푅 represented by t, was expressed as 푟(푡) = 푒−(푡⁄푇) , where the specific time constant T is the time to achieve a 63.2% deposit removal, and R is the slope when using a ‘log (log (1/r))’ vs. ‘log(t)’ plot. The researchers found a 99.5% coefficient of correlation when validating with experimental results using NaOH along with other commercial cleaners to fresh raw milk soiled stainless steel pipes. This proposed model was further examined and interpreted it using the two-factor Weibull distribution. In this manner, the cleaning parameters were expressed as the “lifetime distribution of the soil subjected to a specified cleaning procedure”. Based on the differences in the slope

(slope>1, slope<1 or slope=1), the cleaning process could be further categorized with respect to the changes of cleaning rate. This proposed model along with the unique interpretation was new

46 concept to the deposit removal modeling, however, not many further studies were conducted to confirm or revise this model, under similar soiling and deposit removal process.

Alameda et al. (2011) studied the fatty soil removal from glass surfaces. The deposit consists of a mixture of fatty acids including oleic, palmitic and stearic acids, and nonionic biodegradable surfactants were used in the deposit removal. Mathematical models developed took into account of the soil removal and the simultaneous soil deposition process, and the models were based on the first order kinetic for all the stages involved, and the removal/redeposition process was expressed in equation 2:

′ k1 k → →1 S1 B S2 (2) ← ←′ k2 k2

where S1 is the soiling deposit attached to the substrate, B is the suspended deposit, S2 is the re-deposited soil from the suspended deposit, and k1, k2, k1’ and k2’ are the kinetic constants to be determined from experimental results. The authors defined a detergency of the process, which is the percent of soil in the washing solution (mB) compared to the total initial deposit (m),

m expressed as De(%) = B × 100, and the modelling process, was expressed as De − De = M m

−(k1+k2)t Dem × e , with Dem being the maximum detergency value. When validating the developed model with experimental results, they found good correlation with the model (R2>0.9) indicating an acceptable good of fit.

Other cases extended studies from milk film deposit type to polysaccharide deposit type and study the removal process. Föste et al. (2013) used a starch matrix containing phosphorescent zinc sulfide crystals serving as an optical tracer. From the experimental results, it was found a better cleaning performance when pulsed flow was conducted, as compared to the steady flow

CIP. In another study, a local detection of phosphorescence (LPD) was used (Schöler et al., 2012),

47 and the deposit used was still zinc sulphide layer. The cleaning solution used was either NaOH or water, and the researchers found out a different level of deposit swollen thickness under different treatment. The thickness of the swollen layer after contact with NaOH increased from the dried state of around 50 µm to almost 150 µm at its maximum. With the contact of water, the maximum soil thickness still increased, to a lower extent, of around 40 – 70 µm.

Apart from the milk related biofilm type of deposit and the polysaccharide starch deposit, there were other studies tested the commonly observed microorganisms acquired from a dairy processing pipeline. Lelièvre et al. (2002) started the study of stainless steel pipes with different dimensions in diameter; with characteristics of gradual or sudden contraction or expansion. The deposit used was a strain of Bacillus cereus, isolated from a dairy processing line; and prepared in a staitc condition in water then soiled. In this way, a homogeneous soiling was achieved in the stainless steel pipe inner surfaces. Jullien et al. (2003) tested out different food processing materials for similarities and differences using similar methods. The tested surfaces, however, were preconditioned by immersing into the reconstituted milk at room temperature to acquire a homogeneous biofilm. The materials tested in this study included three grades of stainless steel –

AISI 304, AISI 316, AISI 430, and the surface finish was a pickling finish (2B), a bright annealed finish (2R) or an electropolished finish. The roughness of these surfaces, as well as the surface free energies (determined from the measurement of water contact angle) were measured and compared. These surface topographic characteristic differ a lot under the SEM; the electropolished surfaces were very smooth without any visible scratches or crevices, but on the

2R finish surfaces, more cracks and other surface defects were easily detected (fig. 2-9).

However, from this paper, and other previous studies, no constant conclusions were draw to correlate directly the cleaning performance with the difference in surface finish.

Figure 2-9. SEM images of stainless steel 316 surfaces with different surface finishes of 2B and

2R (Jullien et al., 2003).

Blel et al. (2007) using similar method of the static Bacillus soiling to study the effect of flow arrangement on the spore removal then a more complex pipeline geometries interactions between the spores and the flow eddy scales (Blel et al., 2010). They found out that at low mean shear stress zones, higher levels of adhering spores after the cleaning was present. In addition, the authors restated the importance of the average roughness of the contact surfaces, based on their observation of the high levels of adhering spores in the welding and gasket zones.

Another study used Bacillus cereus spores as the soil and NaOH as cleaning chemical to study the re-adhesion of spores by introducing a loop structure in the pipeline configuration (Le

Gentil et al., 2010). The results showed that after 10 min more than 75% of the initial deposit was removed; after 30 min more between 3-13% was removed. The researchers found that the presence of the loop bend disturbed the flow profile and resulted in a higher level of contamination for the locations downstream. Additionally, given the elongated distance from the inlet, the reduced wall shear stress fluctuations caused a reduced level of spore removal.

Despite the high reproducility of bacterial adhesion soiling and the importance of study specific microorganisms in the dairy processing pipelines, there are drawbacks of using this method. The simple removal of only microorganism in water soiled condition might not truly represent nutritious effect of the milk constituent and the synergistic effect of spore adhesion and milk constituent depositing process. In addition, more than one particular type of microorganisms are commonly found in the dairy process line, and by only studying one single type is far from enough to represent the true soiling condition.

More studies were conducted with respect to physical interpretations to help explain the mechanism of fouling and deposition cleaning. Forces responsible for adhesion between surface and foulant should be overcome to remove the deposit, and these forces include Van der Waals forces, electrostatic forces, hydrogen bonding forces and hydrophobic binding together with contact area effects – the greater the area the greater the total attractive force (Bott, 1995).

Besides forces during the cleaning process, other indicators for the cleaning effects are adhesive strength changes (Liu et al., 2003), heat transfer coefficient changes (Christian and Fryer, 2006), material weight changes (Visser, 1995), deposition layer height changes (Liu et al., 2006a), pressure drop changes (Christian and Fryer, 2006), etc. Among these studies, some of them established simple models for deposits and removal estimation (Liu et al., 2006b; Liu et al.,

2003), but the variables of these models were both physical properties of the deposit layers,

50 without correlating to other surface cleanliness indicators like the indirect ATP bioluminescence method discussed above.

2.5.3 Surface characterization methods

In addition to the experimental observations and computational simulations, more direct detection methods were studied for the milk contact surfaces’ morphology to evaluate the CIP performance. There are many definitions of “clean”: it can be biologically clean, namely no microorganism survived after the CIP process; or chemically clean, namely no residual cleaning chemical remained in the system after the CIP process; or physically clean, most times completed manually and no dirt will be observed visually on the surface; or atomically clean, namely under certain magnification, the characterization of the surface could be determined qualitatively and quantitatively. The atomic level observation should be the most accurate in most of the times, and could be treated as a direct method for the surface characterization due to the high accuracy and sensitivity. Clean should be considered as the system returning to the same condition before milking occurred; in addition, after cleaning, the milk contact surface characteristics should return to the initial state.

From the chemical aspect of the surface, X-ray photoelectron spectroscopy (XPS) or

Fourier transform infrared spectroscopy (FTIR) are suitable options, depending on the specific characteristics of the material surface. By irradiating the specimen sample with a beam of X-rays, the XPS spectra could be obtained. At the same time, the kinetic energy of the X-ray and the number of escaping electrons on the specimen surface could be analyzed. In this way, the specimen surface element composition, concentration and the surface chemical environments could be quantitatively evaluated (Powell and Jablonski, 2009). An FTIR spectrometer collects spectral data in a wide spectral range. In this way, FTIR is able to obtain an infrared spectrum of absorption, emission, photoconductivity or Raman scattering of a solid, liquid or gas. The

51 detection sensitivity of FTRI is 1-2 µm. Therefore, it is more suitable for organic material detection such as aggregation of atoms or functional groups on the surfaces (Powell et al., 2009).

Select options of these surface characterization methods would be developed after several trial experiments.

The optical profilometry is one approach for satisfactorily characterizing the contact surface’s properties. This technique uses an interferometric lens, which moves vertically through the focal plane to focus a beam of white light on the sample tested. The light reflected from the sample surface recombines with a reference beam resulting in the formation of interference fringes. This fringe pattern can be captured on each pixel of a charge-coupled device (CCD) camera array and referenced to the vertical position of the lens. Software packages are available to generate 2D or 3D profiles of the surface (Rousseau et al., 2010). The three dimensional measurements would be sufficient to distinguish the surface characteristics before and after cleaning occurs quantitatively, thus revealing the mechanical force effects of the fluid flow on the material surface. Additionally, the large height (z direction) range (up to 2,000 µm) would help to detect and characterize the deposit thickness and the “corroded” or “pitted” area that result from the use of cleaning solutions.

Scanning electron microscopy (SEM) is another option with high precision. A SEM is able to produce the specimen image by scanning the surface with a focused beam of electrons.

The electrons would interact with electrons residing in the specimen, thus resulting in a variety of signals, which contain the specimen surface topography information and can be detected by certain detector (Banerjee et al., 2009; Jullien et al., 2003). By examining the surface before the soiling, after the soiling, and after the cleaning after a certain CIP process, more information could be acquired as to how the deposit aggregated, swollen, and removed at different stages of soiling and cleaning.

Tissier and Lalande (1986) started to use different visualization methods to study the milk fouling and following cleaning using an experimental size heat exchanger. The SEM results showed, that a densely deposited foulant was found on the top of surfaces right after the treatment started (1 min) and the thickness of the layer increased as time progressed. A more loosely bounded (“spongy”) deposit built up over this dense deposit. The authors used several microscopy methods to observe the topology of the deposit at different stages of the treatments.

Given the treatment of the study being pasteurizing process at high temperature, they observed that the spongy sublayer to be majorly minerals of calcium and phosphate, along with some dispersed fat globules. As the sublayer increased, the percentage of protein in the layer decreased in the spongy deposit. They suspected the required assistance of mineral in building up the initial layer from the chemical microanalysis, however, it was not definite convinced that the initial layer was the mineral layer. Rosmaninho and Melo (2007) used modified simulated milk ultrafiltrate to study the effect of whey protein on calcium phosphate fouling behavior in turbulent flow conditions. On a 2R surface, they found out a spongy like aggregates on the surfaces when looking under the SEM (fig. 2-10). When taking the modified contact surface into account to

+ 2+ reduce the fouling, they found out that out of all the tested modified surfaces (SiF3 , MoS2 and

TiC ion implantation; diamond-like carbon (DLC) sputtering; DLC, DLC–Si–O and SiOx, plasma enhanced chemical vapor Deposition (PECVD); autocatalytic Ni–P–PTFE and silica coating)

(fig. 2-11), Ni-P-PTFE seemed to be the most promising one in reducing the fouling degree and easily cleaned (Rosmaninho et al., 2007). Under lower temperature (44°C), however, the interactions between the protein and calcium phosphate on the surfaces seemed to be different in formation using the simulated milk ultrafiltrate (Rosmaninho and Melo, 2008). They found out that the temperature effect was more distinctive when β-lg was present, possibly due to the inverse relationship between the aggregate type in the liquid and the deposit quantity on the surfaces, when compared to pure mineral solutions (fig. 2-11).

Figure 2-10. Deposition on 2R finish surfaces from modified simulated milk ultrafiltrate solution:

(a) aggregates distribution; (b) calcium phosphate/protein structure (Rosmaninho and

Melo, 2007).

Figure 2-11. SEM micrographs of the mostly amorphous deposits formed from simulated milk

ultrafiltrate at pH = 6.3 and 60°C (flow velocity 0.32 m/s) at various run times on 2R

finish surfaces (Rosmaninho et al., 2007).

2.6 Summary of literature review

Pennsylvania ranks fifth in the United States in milk production from over 7,000 dairy farms. Dairy farming is the leading industry in Pennsylvania, and the state’s economy is highly

54 dependent on dairy farming (Center for Dairy Excellence, 2012). Large profits can accrue from improvements even at a single point of processing on the dairy farm, which benefit not only the farmers, but also the state.

The formation of biofilm in dairy handling and dairy processing equipment poses a great threat to the industry, given the low tolerance of the hazardous microorganisms in the consumed dairy products. The controlling strategy of the biofilm formation is a critically important topic for scientists to investigate; in this context, some advanced research studies are analyzed and compared in the literature review. The cleaning and sanitizing of the dairy processing equipment is essential to guarantee the dairy products’ safety. Therefore, a comprehensive review of the cleaning method for the dairy processing equipment, especially the milking system is also reviewed.

The major emerging cleaning solution utilized in this study, i.e., electrolyzed oxidizing

(EO) water, is thoroughly introduced and explained in the review; including its mechanism and development history, its cleaning and sanitizing effect on produce, other forms of EO water such as near neutral EO water, and the most recent studies in our lab of using EO water as an alternative for milking systems CIP. During processing, fluid milk often encounters fouling from high temperature, and results in an undesirable foulant deposit. The mechanism of milking fouling is explained, along with the possible foulant deposit removal methods involving computational and surface characterization methods. These mechanisms and developed mathematical and computational models are significantly important for this study, providing a starting point to study the non-foulant form of real milk deposit removal during a real milking system CIP; as contrast to the most previous studies of using only sodium hydroxide to chemically remove the constituent-simulated milk deposits.

The first objective of the study is to evaluate the EO water performance for the milking system CIP on a commercial dairy farm, and compare its CIP performance with that of using conventional chemicals. Additionally, conduct a cost analysis between the EO water CIP and 55 conventional CIP (Chapter 3). In the second objective, the novel one-step CIP is studied by using a lab scale pilot milking system using blended EO water by forming different combinations of alkaline and acidic EO water under different solution starting temperature and circulation time duration. Similar comparison with respect to the CIP performance and cost is conducted between the optimal EO water CIP and the commercially available one-step CIP (Chapter 4). In the third objective, an in-depth examination of the CIP process is performed. To that end, a stainless steel surface evaluation simulator is designed and constructed to study the milk deposit removal mechanism during the CIP process. Deposit removal rate models for different CIP cycles are established based on the deposit mass change and validated using the ATP bioluminescence method. Optimized CIP processes with shortened time are established from the resultant models and verified using ATP bioluminescence method (Chapter 5). In the fourth and final objective, deposit removal rate models are developed for the blended EO water one-step CIP and the deposit contact surfaces are studied using SEM (scanning electron microscopy) to examine the morphology of deposits and qualitatively evaluate cleanness of surfaces (Chapter 6).

CHAPTER 3

ELECTROLYZED OXIDIZING WATER FOR CLEANING-IN-

PLACE OF ON-FARM MILKING SYSTEMS – PERFORMANCE

EVALUATION AND ASSESSMENT1

3.1 Abstract

The cleanliness of on-farm milking systems directly affects raw milk quality. A four-step procedure is generally accepted for cleaning-in-place (CIP) of milking systems. The steps are: (1) warm water rinse, (2) alkaline wash, (3) acid rinse, and (4) sanitizer circulation. Electrolyzed oxidizing (EO) water is a novel technology in which acidic EO water and alkaline EO water are generated separately by an electrolyzing weak sodium chloride solution. Because these two solutions match the basic requirements for CIP of milking systems, it was proposed that the EO water can be used as a cleaning and sanitizing agent for CIP of milking systems. Previous studies demonstrated that the utilization of the EO water CIP provided comparable results than the conventional CIP in a pilot-scale milking system. This research was undertaken to evaluate, assess, and validate this technology on a commercial dairy farm and compare it with the conventional CIP. Results from the Adenosine Tri-Phosphate (ATP) bioluminescence method showed that the EO water CIP performance was as good as or better than the conventional CIP for most of the sampling locations and other system components. Moreover, the test for bacterial presence corresponded with the ATP bioluminescence results, indicating the capability of using

EO water for the cleaning and sanitizing of milking systems. For one complete CIP process, it was estimated that the cost of using the EO water CIP is lower than using the conventional CIP by

1 A version of this chapter was published as: Wang, X., Dev, S. R. S., Demirci, A., Graves, R. E., & Puri, V. M. (2013). Electrolyzed Oxidizing Water for Cleaning-In-Place of On-Farm Milking Systems Performance Evaluation and Assessment. Appl. Engr. Agrc., 29(5), 717–726. 57 approximately 25%. Based on the technical performance and economic analyses, this study showed that the EO water had the potential to be adapted as an alternative CIP of milking systems for dairy farms.

3.2 Introduction

According to the 2011 data from the Food and Agriculture Organization (FAO) of the

United Nations, the United States produced 90,865,000 tons of cow milk, ranking first in the world (FAOSTAT, 2012). The annual average consumption of dairy products per capita in the

United States has increased from 244.5 kg in 1975 to 277.6 kg in 2012 (USDA, 2012). The Food and Drug Administration has issued several regulations to guarantee the raw milk quality in the

United States (HHS/FDA, 2007). One of the critical processes in ensuring the raw milk quality is the cleaning process of the milking system. Hence, the cleaning process after every milking should be monitored and controlled with utmost care.

The cleaning-in-place (CIP) process of milking systems usually consists of four steps: warm water rinse, alkaline wash, acid rinse, and sanitizer circulation. The warm water rinse helps to remove the milk residuals from the milking system surfaces. During alkaline wash, by using a high temperature alkaline detergent solution pH of about 11, fat and protein deposits are removed from the system. Acid rinse with a solution pH of about 3 is used to neutralize the alkaline solution, remove the mineral deposits and leave the system in an acidified state to inhibit bacterial growth (table 3-1). The final sanitizer circulation is completed an hour prior to the next milking to destroy any remaining microorganisms (DPC, 2010).

Even though this is a well-accepted process, there is still a need for improvement to make it more efficient, safe, and economical. Electrolyzed oxidizing (EO) water has the potential to achieve these three attributes. EO water is a novel technology in which acidic and alkaline EO

58 waters are generated simultaneously by electrolyzing a weak sodium chloride solution (0.1%).

This occurs in an electrodialysis chamber with a selective membrane between the anode and the cathode (fig. 3-1). Under certain driving voltage and amperage, two solutions are produced; acidic EO water with a pH as low as 2.6, an oxidizing reducing potential (ORP) of 1150 mV, and free chlorine content of 80 ppm and alkaline EO water with a pH of 11.5 and an ORP of -850 mV

(Sharma and Demirci, 2003).

Table 3-1. CIP Recommendations from Dairy Practice Council (DPC, 2010).

Cleaning Cycle Conventional CIP Warm water rinse 2 min; 43.3-48.9°C 8-10 min; start:71.1°C-76.7°C; finish:48.9°C; pH >12.0; 120 ppm Alkaline wash chlorine; 1100 ppm alkalinity; >20 slugs Acid rinse 3-5 min; pH~3.0 Sanitize EPA registered dairy sanitizer solution

Figure 3-1. Mechanism of the EO water generation (Huang et al., 2008).

There have been studies utilizing EO water to effectively clean and sanitize the surfaces of fresh produce and food preparation surfaces. Fabrizio et al. (2005) showed that a 15 s spray with acidic EO water significantly reduced Campylobacter coli on fresh pork surfaces. Bialka et al. (2004) showed the microbial reduction potential of using acidic EO water against E. coli K12 for the egg washing process for a pilot-scale study. Some other studies focused on the effect of using acidic EO water for seafood treatment. Ozer and Demirci (2006) showed that by using the treatment of alkaline EO water followed by acidic EO water on the muscle and skin surfaces of

Escherichia coli O157: H7 and Listeria monocytogenes on inoculated salmon fillets, results are as good as the results of the control group by using 90 ppm chlorine solution treatment.

Despite the progress made on the bactericidal and sanitizing effect that EO water exerted on the products and surfaces, few studies had focused on using EO water to clean and sanitize product processing systems. The properties of acidic and alkaline EO water solutions meet the basic requirements for the CIP of milking system. Using the EO water to clean milking systems has distinct advantages over using the conventional CIP chemicals. The strong chemicals used in the conventional CIP process are potentially hazardous if handled inappropriately; whereas, the

EO water solutions are generated on-site as needed and are not harmful to the operator for a short time exposure. This major advantage motivated researchers to explore the EO water utilization as an alternative to the conventional CIP for milking system. Besides the handling hazards for the operators, the danger posed to the visitors and children on the farm, associated with the storage of strong chemicals is another issue of concern. Moreover, the cost of these chemicals is relatively high compared to that of common salt (NaCl) used for generating the EO water.

Performing CIP by using EO water was first proposed by Walker et al. (2005a and b).

After the evaluation of EO water cleaning performance on individual milking system components, a response surface model was developed to determine the optimal cleaning and sanitizing parameters. The EO water solutions at temperature of 60°C were able to remove all

60 detectable bacteria successfully (Walker et al., 2005a). After determining the optimal solution temperature, they conducted further research by using a pilot-scale milking system to evaluate the short and long term performance of the EO water CIP (Walker et al., 2005b). Results suggested that the cleaning cycle duration of 7.5 min for each step was satisfactory for short term evaluation

(i.e., single CIP cycle). However, they concluded that the cleaning cycle duration of 7.5 min was not satisfactory for the long term evaluation (i.e., up to 10 repeated CIP cycles).

To bridge the knowledge gap of Walker et al. (2005b) study, Dev et al. (2014) conducted further research using the EO water solution for a pilot milking system cleaning. Holding cleaning time duration constant at 10 min, cleaning solution temperatures were further optimized using the response surface method, which suggested that there was no need for heating the acidic

EO water and alkaline EO water to the same temperature of 60°C; using the optimized temperature plus additional safety factor, satisfactory cleaning effectiveness was achieved with starting temperatures for acidic and alkaline EO waters of 45°C and 70°C, respectively.

While encouraging, the pilot-scale milking set-up is a scaled-down system and does not fully and sufficiently represent a commercial-scale dairy farm CIP; therefore, this study was conducted on a commercial dairy farm to evaluate the EO water performance for CIP of the milking system using the above mentioned optimized parameters.

3.3 Materials and Methods

3.3.1 Preparation of the EO water

EO water solutions were generated with an EO water generator (Model ROX60SA,

Hoshizaki Electric Co. Ltd, Sakae, Toyoake, Aichi, Japan). Sodium chloride solution was generated automatically by filling salt in the front chamber. By adjusting the voltage and amperage of the generator, the acidic EO water had (1) pH of about 2.6, (2) ORP of about 1150

61 mV, (3) free chlorine content of about 80 ppm, whereas the pH of the alkaline EO water was 11.5 with an ORP of -850 mV. The pH and ORP of both solutions were tested by using a pH/ORP meter (Model 445, Corning, Inc., Big Flats, N.Y.), and chlorine content of the acidic EO water was tested by titrating against an N, N-diethyl-p-phenylenediamine-ferrous ethylene diammonium sulfate (DPD-FEAS) solution prepared using a test kit (Hach, Inc., Loveland, Colo.). Alkaline EO water was heated using a 304.8 L (80 gal) capacity tank water heater (Model RUE PRO-80-2,

Ruud Manufacturing Co., Atlanta, Ga.) and acidic EO water was heated using a tankless heater

(Model EX1608TC, Eemax Inc., Oxford, Conn.). The schematic set-up of all these facilities is shown in figure 3-2.

Figure 3-2. Overall schematic of the EO water generator and other units for milking system CIP

on a commercial dairy farm and associated flow loops (light grey loop: acidic EO water;

dark grey loop: alkaline EO water; black loop: water supply).

3.3.2 Farm Trial

A mid-size commercial dairy farm with 81 cows was used for the study. This dairy farm had a tie stall barn with around the barn pipeline milking system with a pipeline length of about

140 m (460 ft) and milk quality that ranked above normal; i.e., the farm represented a typical commercial dairy farm in Central Pennsylvania. This trial was conducted with the concurrence and approval of Pennsylvania Department of Agriculture.

The 4-month evaluation time frame was divided into three periods: (1) first month for the conventional milking system CIP performance testing and to establish the baseline controls, (2) second and third months for the EO water milking system CIP performance testing, and (3) fourth month for monitoring of the conventional CIP process to ensure that no potential hazard was brought into the milking system as a result of this study. Average data were obtained during 1st and 4th periods for conventional CIP, whereas average data were obtained during 2nd and 3rd periods for EO water CIP.

A detailed comparison of cleaning conditions between the EO water and conventional milking system CIP is listed in table 3-2. In general, during the CIP process, sufficient vacuum pump capacity for air-injected wash was provided to achieve the slug requirement (cleaning solutions circulating for about 8 to 10 min with a minimum of 20 slugs) (DPC, 2006). A 100 L

(27 gal) cleaning solution was used during each cleaning cycle. However, the major differences between these two cleaning methods are: (1) Two warm water rinses with a starting temperature of approximately 37°C were used for the EO water CIP, namely before and after alkaline wash cycle. For the conventional CIP, only one warm water rinse cycle before alkaline wash was utilized. The second warm water rinse used in the EO water CIP is recommended by Dairy

Practices Council (DPC, 2006) and it would increase the efficiency of acidic EO water by preventing neutralization of the alkaline EO water. (2) Alkaline EO water solution was heated to about 70°C, and acidic EO water solution was heated to about 40°C based on the study of Dev et 63 al. (2014). For the conventional CIP, the alkaline cleaning solution was heated to about 65°C and unheated (room temperature) acid rinse solution was used.

Table 3-2. Cleaning condition comparisons per cycle between the EO water CIP and the

conventional CIP on a commercial dairy farm.

Cleaning Cycle EO Water CIP Conventional CIP 5 min 5 min Warm water rinse Start: 37°C Start: 37°C 10 min 10 min Start: 72°C Start: 65°C Alkaline wash Finish: 46°C Finish: 42°C pH 11.5 pH 11.5 80 ppm chlorine 125 ppm chlorine 5 min Warm water rinse N/A[a] 37°C 8 min 7.5 min Start: 40°C Start: 18°C Acid rinse Finish: 20°C Finish: 15°C pH 2.6 pH 2.5 [a]Not applicable.

The conventional CIP chemicals used on this farm were: sodium hydroxide/sodium hypochlorite mixture of alkaline solution (Liquid Pfite, GEA WestfaliaSurge Inc., Naperville, IL) and phosphoric/sulfuric blend of acid solution (Dairy Star CIP Acid Cleaner, distributed by GEA

WestfaliaSurge Inc.).

3.3.3 Sampling Sites

In order to gather sufficient data for robust statistical analyses, nine sampling locations at tri-clamp connections along the stainless steel system pipelines were chosen. These sampling locations were elbow and straight pipeline connections including the gasket used at these connections; which facilitated the sampling procedure and reduced the labor. These sampling locations are identified as A through I (fig. 3-3).

These nine sampling locations along the system pipelines can be divided into three categories: Locations A and B both had 45° elbows with vacuum and slugs during the cleaning process; Locations C, D, and E had 90° elbows with vacuum and slugs during the cleaning process; the rest of the locations F, G, H, and I had 90° elbows but did not experience vacuum or slugs during the cleaning process, as these were pipeline connections after the transfer pump in the milking system. In addition to these sampling locations, other system components such as liners, milk hoses, and milk inlets were sampled.

Figure 3-3. (a) General overview schematic of farm milking and cleaning system; (b) detailed

sampling locations along the pipelines in the milk house.

3.3.4 Sampling Protocols

The nine sampling locations and other system components of liners, milk hoses, and milk inlets were analyzed for Adenosine Triphosphate (ATP) bioluminescence and bacterial presence.

Samples were collected from the nine sampling locations A through I, by disassembling the connections and swabbing the inner surfaces of elbows and straight pipelines. Gaskets in tri- clamp locations were sampled as well (fig. 3-4).

With reference to the cleaning solution flow direction, the right half of the connections of straight pipelines, elbows, and gaskets were sampled using ATP swabs for bioluminescence analyses and the left half using sterile calcium alginate-tipped applicators for analyzing the bacterial presence. Other system components of liners, milk hoses, and milk inlets were swabbed over the entire circle instead of half circle due to the limited availability of milk contact surfaces for swabbing in these components. On the sampling day, three randomly assigned locations out of nine were sampled following the sequence of A, F, H; B, C, E, and D, G, I. For example, on Day

1, locations A, F, and H were sampled; on Day 2, locations B, C, and E were sampled; and on

Day 3, locations D, G, and I were sampled. In this way, three ATP bioluminescence and bacterial presence data of elbows, three ATP bioluminescence and bacterial presence data of straight pipelines and three ATP bioluminescence and bacterial presence data of gaskets were collected on one sampling day. Meanwhile, three liners, three milk hoses and three milk inlets were sampled for ATP bioluminescence analyses and the other three of each category were sampled for bacterial presence analyses on the sampling day.

Figure 3-4. Sampling schematic detail: (a) elbow; (b) straight pipeline.

3.3.5 Sample Analyses

Inner surfaces of all sampling sites were sampled using PocketSwab Plus swabs (Charm

Science, Inc., Lawrence, MA) for ATP bioluminescence analyses and sterile calcium alginate- tipped applicators (Puritan Medical Production Co. LLC, Guilford, Maine) for analyses of bacterial presence. The ATP swabs’ firefly enzyme luciferase gave a quantitative measurement in terms of Relative Light Unit (RLU) by using a novaLUM palm-sized luminometer (Charm

Science, Inc.); the RLU readings served as an indicator of the surface cleanliness. Bacterial samples were collected with sterile calcium alginate-tipped applicators, which were placed in a

Tryptic Soy Broth medium and incubated at 37°C for 48 h for bacterial enrichment. The clear medium observed visually was accepted as “negative” while the opaque (turbid) ones were accepted as “positive.” For further discussion purposes, negative enrichment percentage was calculated by percentage of the number of negative samples with respect to the total number of samples (number of negative samples and positive samples) for sampling locations and other system components.

3.3.6 Cost Analyses

Previous studies predicted that using EO water for milking system CIP would be economical when compared to the conventional CIP as long as cost recovery of the EO water generator is not considered (Walker, 2003). However, the detailed cost comparison between these two methods was not done. Herein, it was proposed that the additional cost analyses included in this study would further prove the cost effectiveness of using the EO water CIP of milking systems. However, due to the lack of available information needed for such assessments, several assumptions were made in the calculations and analyses.

1. Based on the local average temperature during the evaluation time frame, it was assumed that the local water supply temperature was constant at 10°C and the temperature was not affected by the changing ambient temperature.

2. Rinse water and alkaline wash solution for both CIP methods were heated and stored using an insulated tank heater, and it was assumed that the heat loss during storage was negligible. Acidic EO water was heated using a tankless heater, and it was assumed that the electrical energy was fully converted into thermal energy. In addition, it was assumed that negligible heat loss occurred during the transport of heated solution.

3. The concentrations of both acidic and alkaline EO water cleaning solutions were

0.12% and the concentration of the conventional cleaning chemical solution was 0.6%. Therefore, it was assumed that the specific heat of rinse water, alkaline wash solution and acid rinse solution were the same as pure water for the convenience of calculation.

4. Conventional chemical solution usage was handled by the farmer, and it was assumed that the amount of chemical solution usage for each CIP was constant.

5. Water supply was from the farm well powered by a jet pump. The well was buried below ground level and specifications were unknown; therefore, the power used to run the pump was estimated. Based primarily on the number of the cows on farm, a one-half horsepower cast iron convertible well jet pump was used for calculation.

6. The cleaning processes for both the conventional CIP and EO water CIP systems were fully automated. Maintenance and labor cost of the milking system using both the conventional

CIP and the EO water CIP were assumed same for the milking system configurations. Moreover, the maintenance cost for EO water generator was lack of sufficient information during the farm trial period to distinguish a significant difference between conventional CIP and EO water CIP.

Therefore, the maintenance or labor cost was considered to be negligible for this study.

i. Cleaning Chemicals

Chemical for EO water generation was sodium chloride. Chemicals for the conventional

CIP were used as indicated above (Farm Trial section). The total cost for each cleaning cycle for the conventional CIP was calculated based on the amount of chemicals required after factoring in the dilution.

ii. Cleaning Protocol Differences

As stated above, the EO water CIP used warm water rinses both before and after alkaline washing cycle; however, for the farm conventional CIP, the second warm water rinse was cut to

69 reduce heating and mechanical usage (table 3-1). For the EO water CIP, the heated acidic EO water was used for the acid rinse cycle, and for the conventional CIP, the unheated acid cleaning solution was used after alkaline wash cycle.

In addition, the farm conventional CIP did not include the sanitizer circulation prior to the next milking, which reduced the total cost accordingly. However, as mentioned above, the acidic

EO water functions both as a cleaning and sanitizing agent, which means the acid rinse cycle and the following sanitizer circulation were combined into one cycle when using the EO water CIP.

Herein, the scenario of using acid sanitizer (chemical used for the last cleaning cycle functions for both cleaning and sanitizing) instead of acid rinse chemical for the last CIP cycle was also calculated to assist further analysis.

iii. Operational Cost Analyses

The operational cost of one CIP process was calculated based on the following analyses.

Material and energy consumptions during the cleaning cycle were taken into consideration. These included: (1) salt to generate the EO water and salt to soften supply water thus reducing the supply water hardness, (2) conventional alkaline and acid chemicals, (3) water usage for each cycle, and (4) electricity usage which includes the solution heating and mechanical consumption for each cycle of one complete CIP process. The general comparison is listed in table 3-3.

Table 3-3. Operational cost comparison between the EO water and conventional CIP on a

commercial dairy farm.

EO Water CIP Cost ($) Percent (%) Conventional CIP Cost ($) Percent (%)

Cleaning agent 0.21 9.7 Cleaning agent 1.80 63.4

Heating usage 1.64 76.3 Heating usage 0.89 31.3

Mechanical usage 0.30 14.0 Mechanical usage 0.15 5.3

Total 2.15 100 Total 2.84 100

Essential equations used for calculations:

Q = Cp×m×Δt (Q: heat; Cp: Specific heat; m: mass; Δt: temperature increment)

W = V×A (W: Power; V: Voltage; A: Amperage)

3.4 Results and Discussions

The four-month on-farm study demonstrated the capability of EO water for CIP of the

milking system. Based on the testing methods described above, ATP relative light units (RLU)

readings and microbial presence analyses are performed to evaluate the effectiveness of the EO

water CIP compared with the conventional CIP. The operational cost comparison between EO

water CIP process and the conventional CIP process is calculated for further analyses.

3.4.1 ATP RLU Readings

Due to the high sensitivity of the test, the RLU readings of ATP tests vary from zero to

millions depending on the surface soiling status. The RLU reading of zero represents the surface

to be clean and higher RLU readings represented “dirtier” surfaces. The manufacture -

71 recommended practical cut-off of RLU reading in determining the surface cleanliness is 1,000 or below for stainless steel (namely elbow, straight pipeline, and milk inlet in this study) and 4,500 or below for porous rubber materials (liner and milk hose). A logarithm transformation was used to adjust the wide-ranging results to be normally distributed for better statistical analyses (fig. 3-

5). By using “RLU+1” the indeterminate value for log of zero was avoided. Furthermore, addition of 1 RLU reading to the actual measured value has negligible effect compared to the variations in measured values on the logarithm RLU+1 and the results used for comparison.

Figure 3-5. Average RLU reading comparison between the EO water CIP and the conventional

CIP: (a) elbow, (b) straight pipeline, (c) gasket, and (d) liner, milk hose, and milk inlet.

Each sampling location includes about eight replications.

The average RLU readings of most sampling sites, including sampling locations and system components, were lower than the cut-off RLU readings for both the EO water CIP or the conventional CIP. This indicated that both the CIP methods were performing without introducing or leaving any residual potential hazards to the milk quality during the evaluation process. For each sampling location and system component, RLU readings were summarized as follows:

i) Elbows

Most of the elbow average RLU readings were below the cut-off value, while using both methods except for the location H (fig. 3-5a). However, the EO water average RLU readings were lower than the conventional average RLU readings. For example, for the conventional and EO methods at the elbow of sampling location G, the average log (RLU+1) readings were 1.19 and

0.51, respectively.

A comparison between different categories showed that the average RLU readings of elbows were generally lower than other milking system components. One of the reasons is that the elbows are made of smooth stainless steel and it is not easy for microorganisms and other deposits to attach onto the surface, thus making it easy to be cleaned by the solution. Another important reason is the curvature of elbowseither a 45 or 90° bend, which would result in additional dynamic forces due to the change in direction of the fluid flow over and above the static pressure force and frictional forces; this facilitates better flow turbulence and recirculation thereby enhancing the cleaning performance (Bansal, 2005).

ii) Straight Pipelines

In most of the straight pipelines, the average RLU readings were below the cut-off value for both methods, except for the location H (fig. 3-5b). The EO water average RLU readings were lower than the conventional average RLU readings. For example, for the conventional and EO

73 methods at the straight pipeline of sampling location C, the average log (RLU+1) readings were

1.60 and 1.54, respectively. Similarly, the stainless steel material allowed straight pipelines to be cleaned easily, resulting in lower average RLU readings compared with other system components and sampling locations.

iii) Gaskets

Most of the gasket average RLU readings were below the cut-off value for both methods, again except for the location H (fig. 3-5c). In general, EO water average RLU readings were lower than the conventional average RLU readings. For example, for the conventional and EO methods at the gasket of sampling location I, the average log (RLU+1) readings were 3.27 and

3.09, respectively.

On the other hand, for all nine sampling locations, the gasket average RLU readings were higher in general than the elbow and straight pipeline average RLU readings. This can be attributed to the material of the gaskets. The material of gasket was made of rubber; Gaspar-Rolle

(1991) showed that the tough topography with “caverns” and crevices spread over the surface of buna-N-rubber resulted in greater numbers of Pseudomonas fragi, Listeria monocytogenes Scott

A and Bacillus cereus attachment. In addition, any pitting due to the use of powerful cleaning solution could make the scenario worse.

iv) Liners, Milk Hoses, and Milk Inlets

As is shown in figure 3-5d, the average RLU reading for the liner when using the EO water CIP was significantly lower (P<<0.05) than the conventional CIP; log (RLU+1) values were 0.09 and 0.78, respectively. The average RLU reading for the milk hose when using the EO water CIP was lower than the conventional CIP, but not statistically different (P>0.05). The average RLU reading for the milk inlet when using the EO water CIP was higher than the conventional CIP, but readings from both methods exceeded the cut-off value of 1,000. This was 74 caused by the fact that milk inlets along the pipeline on this farm were placed at an angle towards the ceiling, which made it very difficult for the milk inlet inner surface to be cleaned thoroughly as a result of the deceleration by gravity of the cleaning solution, even with the vacuum and slugs.

v) Location H

As mentioned in preceding subsections, average RLU readings were always high for both the EO water CIP and the conventional CIP at location H. An explanation for these high average

RLU readings comes from the configuration of this location. Location H had neither vacuum nor slugs during the cleaning process. When the cleaning process is finished and the vacuum shuts down, cleaning solution would remain in the vertical pipeline before being drained through the transfer pump. This resulted in location H having the leftover cleaning solution. With time, the cleaning waste floating on the top left the location H “dirty” for both the conventional CIP and

EO water CIP methods. But the overall average RLU readings for the EO water CIP were less than those of the conventional CIP even at this location.

In conclusion, the average RLU readings for the EO water CIP were lower than the conventional CIP for most sampling sites. The performance difference was not significantly different between the EO water CIP and the conventional CIP (p>0.05), and therefore, the EO water CIP performance was statistically as good as the conventional CIP performance in terms of average RLU readings, because there were no significant differences between conventional CIP and EO water CIP for the sampling locations (Table 3-4).

Table 3-4. p-values between the conventional CIP and EO water CIP ATP bioluminescence

evaluations for the sampling locations of straight pipelines, elbows, gaskets, liners, milk

hoses, and milk inlets.

Sampling locations Straight pipelines Elbows Gaskets Liners Milkhoses Milk inlets

A 0.362 0.966 0.098 0.000 0.108 0.447

B 0.508 0.067 0.592

C 0.926 0.180 0.636

D 0.407 0.775 0.364

E 0.321 0.916 0.154

F 0.106 0.087 0.409

G 0.773 0.312 0.285

H 0.341 0.402 0.014

I 0.733 0.648 0.813

3.4.2 Microbiological Enrichment

In order to evaluate sanitation effectiveness, microbiological analysis was conducted by using simple enrichment protocol. Negative enrichment percentage was calculated by using negative sample number divided by the total number of samples for each sampling site; higher negative enrichment represents less microorganism growth. Per the plan of the sampling procedure, not all of these nine sampling locations were sampled the same number of times (fig.

3-6).

i) Elbows

Most of the elbows had higher negative enrichment percentages when using the EO water

CIP compared with the conventional CIP. For example, for the conventional and EO methods at sampling location B, the negative enrichment percentages were 67% and 100%, respectively.

Other examples are locations A and E, whose negative enrichment percentages of using the EO water CIP were 100%, which were 10% or higher than the corresponding conventional CIP negative enrichment percentages.

Figure 3-6. Negative bacterial enrichment comparison between the EO water CIP and the

conventional CIP: (a) elbow, (b) straight pipeline, (c) gasket, and (d) liner, milk hose, and

milk inlet. Each sampling location includes about eight replications.

Regarding location H, the bacterial enrichment showed similar results like ATP average

RLU reading. The negative enrichment percentage for this elbow was much lower than that of other sampling locations, which were 38% for both the conventional CIP and the EO water CIP.

This result also indicated that at this particular location, irrespective of the type of solution used, effective cleaning could not be achieved to prevent the microorganisms from growing due to the fact explained above.

i) Straight Pipelines

Most of the straight pipelines had higher negative enrichment percentages when using the

EO water CIP compared with the conventional CIP method. For example, for the conventional and EO methods at sampling location A, the negative enrichment percentages were 78% and

100%, respectively. Similarly, negative enrichment percentages for location H were 0% for both the conventional CIP and the EO water CIP. Obviously, this problematic configuration needs to be addressed in further studies to prevent any hazard into the milking system.

ii) Gaskets

For gaskets, the negative enrichment percentages of the EO water CIP were as good as the conventional CIP. For example, for the conventional and EO methods at sampling location A, the negative enrichment percentages were 44% and 75%, respectively, when comparing the negative enrichment percentages among elbows, straight pipelines and gaskets, the negative enrichment percentages of gaskets were lower in comparison. For both location H and location I, the negative enrichment percentages were 0%. This result corresponded with the average RLU readings of these two sampling locations, indicating the pronounced effect of surface material and location configuration on the cleaning performance in the milking system.

iii) Liners, Milk Hoses, and Milk Inlets

For liners, milk hoses, and milk inlets, the negative enrichment percentages of the EO water CIP were higher than the conventional CIP. For example, for the conventional and EO methods of milk hose, the negative enrichment percentage were 5% and 8%, respectively. These results also corresponded to the average RLU readings. Microorganisms grew more actively within the surfaces of milk hoses and milk inlets; due to the fact that the material of milk hose is porous and the anti-gravity direction of milk inlet setting was not conducive for cleaning as discussed above. Higher negative percentages for liners resulted from the immersion of the liners in the cleaning solutions for longer time duration. In addition, during warm water rinse cycle

(which is really important to the cleaning performance), rinse water coming back from the system pipelines was drained directly instead of re-circulating and re-contaminating within the liners

(during alkaline wash and acid rinse cycle, the cleaning solutions were re-circulated). This is another reason for the liners to be cleaner.

In conclusion, the analyses for bacterial presence also suggested that the EO water CIP performance was comparable to the conventional CIP method.

3.4.3 Cost Comparison

The analyses of cost for each CIP method would help to compare and show the advantage of using EO water for CIP of milking systems in addition to the environmental benefits and decreased potential chemical hazards. Table 3-3 shows the three major portions of the cost: cleaning agents, heating cost, and mechanical usages that include the usages of air injector and various pumps, etc. In general, the cost for the EO water CIP per cycle was calculated as $2.15, whereas the cost for the conventional CIP per cycle was calculated as $2.84. Therefore, using the

EO water CIP was about 25% less expensive than using the conventional CIP of milking system.

When comparing the cleaning agent costs for both methods, using the EO water CIP showed an apparent advantage over using the conventional CIP. The cleaning agent cost of conventional chemical was almost nine times higher than that of EO water solutions; conventional chemical cost was $1.80 per cycle and the cost of salt for generating EO water solutions was $0.21 per cycle.

It had been reported that for some dairy farms, heating water accounts for about 25% of the total energy used on farm (Sanford, 2003). This is true even specifically for milking system cleaning process; heating water did take a great percentage in the total cost (76.3% for the EO water CIP and 31.3% for the conventional CIP). Reducing the temperature of each cleaning solution would definitely lower the total cost significantly, which might explain why the conventional CIP had only one warm water rinse and used unheated acid solution.

In addition, for each CIP process, current calculations were based on the farm practice. If the conventional method followed the same cleaning procedure of the EO water CIP, namely adding an additional warm water rinse cycle after alkaline wash, and used heated acid solution, an additional $0.72 of water heating usage would be added to the total cost of the conventional CIP process. Moreover, if the conventional CIP was conducted with the sanitizer circulation prior to the next milking as recommended, an estimated additional $0.11 would be added to the total cost of the conventional CIP process; based on one of the sanitizer advertisements posted online (GEA

Farm Technologies, 2012). Together these two items would result in an add-on of $0.83 to the total cost of the conventional CIP. In that case, the total cost of the conventional CIP would be

$3.67, resulting EO water to be about 40% less expensive than using the conventional CIP of milking system.

However, the capital cost must be thoroughly considered for the EO water generator.

Currently the expenditure of this machine might be difficult for most farmers to afford. However, the retail price might be lowered due to scale of economies rule if the demand is higher. 80

3.5 Conclusions

This study, conducted on a commercial dairy farm, was aimed to evaluate and assess the cleaning performance between the EO water CIP and the conventional CIP of milking systems.

Based on the four-month trial, the EO water achieved cleaning effectiveness comparable with the conventional CIP method on this representative commercial dairy farm for most sampling locations and milking system components, by using ATP bioluminescence and bacterial enrichment methods. From the cost analyses aspect, it was estimated that the cost of using the EO water CIP is lower than using the conventional CIP by approximately 25%, provided an EO water generator unit is already purchased.

To further promote and develop this technology of using EO water for CIP of milking system, more research is required to study its long term performance and its adaptability to various dairy farm configurations. With regard to the cost analyses, further study should focus on the optimization of the cleaning parameters for each cleaning cycle, including the starting temperature, pH of alkaline wash solution and acid rinse solution, appropriate chlorine concentration, proper cleaning duration for each cycle, etc. The goal should be to lower the solution temperature while maintaining similar or better cleaning effectiveness when compared with the conventional CIP method.

In conclusion, EO water is shown to be a promising technology for CIP of milking system, both from cleaning/sanitizing performance aspect and from the economic aspect.

Previous studies of using EO water to clean product surfaces laid a solid scientific foundation, but the current study had further successfully expanded the application of EO water for CIP of a commercial scale dairy farm.

3.6 Acknowledgements

Funding for this project was provided in-part by a USDA Special Research Grant (No.

2010-34163-21179) and the Pennsylvania Agricultural Experiment Station. We are also thankful to Hoshizaki Electric Co. Ltd. (Sakae, Toyoake, Aichi, Japan) for the technical support for the EO water generator used in this study. We also would like to acknowledge Tim Peachey, Hasan

Coban, Roderick Thomas, Randall Bock, and Robin Pritz for their help in the project.

3.7 References

Bialka, K. L., Demirci, A., Knabel, S. J., Patterson, P. H., & Puri, V. M. (2004). Efficacy of

electrolyzed oxidizing water for the microbial safety and quality of eggs. Poultry Sci.,

83(12), 2071-2078.

Bansal, R. K. (2005). A Text Book of Fluid Mechanics and Hydraulic Machines, (pp. 286-292) New

Delhi, India.: Laxmi Publications.

Dev, S. R. S., Demirci, A., & Graves, R. (2011). Optimization and modeling of an electrolyzed

oxidizing water based clean-in-place technique for farm milking systems using a pilot-

scale milking system. J. Food Engr., 135, 1-10.

DPC. 2006. Guideline for effective installation, cleaning and sanitizing of tie barn milking systems.

Dairy Practices Council Publication Number 102, April. Richboro, Pa.: DPC.

DPC. 2010. Guidelines for installation, cleaning, and sanitizing of large and multiple receiver parlor

milking systems. Dairy Practices Council Publication Number 4, February. Richboro, Pa.:

DPC.

Fabrizio, K. A., Sharma, R. R., Demirci, A., & Cutter, C. N. (2002). Comparison of electrolyzed

oxidizing water with various antimicrobial interventions to reduce Salmonella species on

poultry. Poultry Sci., 81(10), 1598-1605.

Fabrizio, K. A., & Cutter, C. N. (2005). Application of electrolyzed oxidizing water to reduce

Listeria monocytogenes on ready-to-eat meats. Meat Sci., 71(2), 327-333.

FAOSTAT. (2012). Total production of cow milk, whole fresh cow. Retrieved from

http://faostat3.fao.org/faostat-gateway/go/to/browse/Q/QL/E.

Fasenko, G. M., Christopher, E. E. O., & McMullen, L. M. (2009). Spraying hatching eggs with

electrolyzed oxidizing water reduces eggshell microbial load without compromising broiler

production parameters. Poultry Sci., 88(5), 1121-1127.

Gaspar-Rolle, M. N. P. (1991). Attachment of bacteria of Teflon ad buna-N-rubber gasket materials.

Ph.D. Dissertation. Blacksburg, Va.: Virginia Polytechnic Institute and State University.

GEA Farm Technologies. (2012). Retrieved from http://www.gea-

farmtechnologies.com/us//en/bu/farm_services/hygiene/equipment_facility/cip/cip_disinfec

tion/zinicin/default.aspx.

HHS/FDA. (2007). Guidance for industry: Dairy farms, bulk milk transporters, bulk milk transfer

stations and fluid milk processors: food security preventive measures guidance.

Washington, D.C.: HHS/FDA.

Huang, Y. R., Hung, Y. C., Hsu, S. Y., Huang, Y. W., & Hwang, D. F. (2008). Application of

electrolyzed water in the food industry. Food Cont.,19(4), 329-345.

Ozer, N. P., & Demirci, A. (2006). Electrolyzed oxidizing water for decontamination of raw salmon

inoculated with Escherichia coli O157:H7 and Listeria monocytogenes Scott A and

response surface modeling. J. Food Engr., 72(3), 234-241.

Sanford, S. (2003). Heating water on dairy farms. University of Wisconsin – Cooperative Extension

Publication A3784-2. Madison, Wis.: University of Wisconsin.

Sharma, R. R., & Demirci, A. (2003). Treatment of Escherichia coli O 157: H7 inoculated alfalfa

seeds and sprouts with electrolyzed oxidizing water. Int. J. Food Microbiol., 86(3), 231-

237.

USDA. (2012). USDA Economic Research Service - Dairy Data. Retrieved August 13, 2014, from

http://www.ers.usda.gov/data-products/dairy-data.aspx#.Uv0x2vldV8E,

Walker, S. P. (2003). Cleaning milking systems using electrolyzed oxidizing water. MS Thesis.

University Park, Pa.: Pennsylvania State University, Department of Agricultural and

Biological Engineering.

Walker, S. P., Demirci, A., Graves, R. E., Spencer, S. B., & Roberts, R. F. (2005a). Response surface

modeling for cleaning and disinfecting materials used in milking systems with electrolyzed

oxidizing water. Int. J. Dairy Technol., 58(2), 65-73.

Walker, S. P., Demirci, A., Graves, R. E., Spencer, S. B., & Roberts, R. F. (2005b). Cleaning

milking systems using electrolyzed oxidizing water. Trans. ASAE, 48(5), 1827-1833.

CHAPTER 4

EVALUATION OF BLENDED ELECTROLYZED OXIDIZING

WATER-BASED CLEANING-IN-PLACE TECHNIQUE USING A LAB

SCALE PILOT MILKING SYSTEM

4.1 Abstract

Milk safety is a food safety concern in the United States. The cleanliness of on-farm milking systems directly affects raw milk quality. The conventionally-accepted four-step procedure for milking system Cleaning-in-Place (CIP) includes: (1) warm water rinse, (2) alkaline wash, (3) acid wash, and (4) sanitizing rinse prior to the next milking event. Electrolyzed oxidizing (EO) water is an emerging technology, which generates acidic and alkaline EO water by the electrodialysis of dilute sodium chloride solution. Previous studies had shown that EO water can be an alternative for conventional milking system CIP, both in pilot scale milking systems and on a commercial dairy farm. Recently, a one-step cleaning process has been adopted on increasing numbers of dairy farms, which combines the alkaline wash and acid wash together to save chemical expenditure, water usage, energy cost, and time. By blending acidic EO water with alkaline EO water, a less corrosive, but still effective blended EO water solution can be produced.

Therefore, this study was undertaken to evaluate the blended EO water solution for one-step CIP using a pilot milking system and compare the CIP effectiveness and economical cost with commercial one-step CIP chemicals. Box-Behnken three-factor response surface method was used to determine the optimal cleaning time, the starting temperature of the blended EO water solution, and the acidic EO water percentage. Also, two commercial one-step CIP chemicals were used for CIP effectiveness and economical cost comparisons. Results showed that a cleaning time of 17 min, a starting temperature of 59°C and an acidic EO water percentage of 60% in the 85 blended EO water solution could achieve the required 100% CIP performance and was comparable to the commercial one-step cleaning chemicals. Moreover, it was determined that for one complete CIP process the operational cost of using blended EO water was lower by 80% than using the commercial one-step cleaning chemicals. Overall, this study demonstrated that the blended EO water had the potential to be adapted as an alternative for one-step CIP for the milking systems and possibility of other food processing equipment.

4.2 Introduction

The United States ranks first in the world of raw milk production and high consumption

(FAOSTAT, 2012). Quality and safety of consumed milk is always important, therefore, the Food and Drug Administration has issued regulations including the collection, on-farm storage, cleaning and sanitization, transportation, milk house daily maintenance and so on to assure the raw milk quality and safety in the United States (FDA, 2012). It is of great importance to ensure the safety of raw milk, as it is one of the key steps to ensure the dairy product safety and provide the public with safe milk for direct consumption.

Cleaning-in-place (CIP) of the milking system is performed right after the milking event, and conventionally it consists of four steps: warm water rinse, alkaline wash, acid wash, and a sanitizing rinse prior to the next milking event. EO water is an emerging technology using electric current in an electrodialysis chamber to generate alkaline and acidic EO water solutions (Chapter

3).

Studies had been conducted in our lab using EO water to clean the milking system CIP; using milking system related materials, and a lab scale pilot milking system (Walker et al., 2005a;

Walker et al., 2005 b; Dev et al., 2014). A further CIP performance comparison of using EO water on a dairy farm was presented in Chapter 3 and its efficacy and cost was compared. Based on a

86 four-month trial on a mid-size commercial dairy farm, it was concluded that the EO water achieved similar cleaning effectiveness compared to the conventional CIP method for most sampling locations and milking system components. In addition, based on the cost comparison of operational analysis, the estimated operational cost of using the EO water CIP was lower than that of using the conventional CIP by approximately 25%, provided an EO water generator unit and other associated accessories are already in place. EO water is therefore shown to be a promising technology for the CIP of milking systems, both from cleaning/sanitizing performance and economic aspects.

Recently, a new CIP approach has received attention on increasing numbers of dairy farms (Parr, 2013). This new CIP approach combines the alkaline wash and the acid wash cycles into one, to conduct a “one-step” CIP for the milking system. Several commercially available one-step CIP chemicals are on the market and their CIP performance is claimed to be comparable to the conventional separate alkaline wash and acid wash of the milking system. It was claimed that using one-step CIP can be cost saving through eliminating the separate alkaline and acid wash cycles, decreasing the usage of water, chemicals, energy, and time. In addition, the manufactures claimed that their products are capable of keeping the bacterial count low for standard tests. However, there aren’t any published experimental results for the milking system

CIP using these chemicals.

There have been some commercially available EO water generators that produce a mixed

EO water solution, named as “near neutral EO water” or “mixed oxidant”. The mixing of acidic

EO water into alkaline EO water still contains chlorine, but possesses a relatively lower corrosive pH, while having a high ORP to function as a disinfecting agent (Guentzel et al., 2008).

Researchers assessed the disinfecting effectiveness of near neutral EO water for the inactivation of five microorganisms (Escherichia coli, Salmonella Typhimurium, Staphylococcus aureus,

Listeria monocytogenes, and Enterococcus faecalis) on the surfaces of spinach and lettuce, and

87 results showed a 4.0 – 5.0 log10 CFU/ml bacterial reduction for all five microorganisms (Guentzel et al., 2008). In addition, compared to a water rinse, the near neutral EO water treatment of spinach and lettuce surfaces does not leave any significant residual chlorine. The effective utilization of near neutral EO water on peach and grape surfaces postharvest sanitation to prevent incidence of microorganism infection (Botrytis cinerea and Monilinia fructicola) was also demonstrated (Guentzel et al., 2010). These results expand the application of near neutral EO water from simple sanitization onsite to the enhancement of fruit shelf life in commercial markets. When comparing to other fungicide like “Captan” to conduct the disinfecting experiments on strawberry plants, study showed that using near neutral EO water with 100 ppm chlorine concentration for the strawberry plant spray did not leave significant phytotoxicity compared to the water treatment, which indicates that near neutral EO water treatment is truly promising from a operation safety aspect (Guentzel et al., 2011). Given the rapid development of near neutral EO water applications and the desired one-step milking system CIP approach, this study was undertaken to evaluate the blended EO water CIP process for the pilot milking system as a one-step milking system CIP alternative.

4.3 Materials and methods

4.3.1 Microorganisms and Medium

A cocktail of four types of commonly found microorganisms in raw milk, including

Pseudomonas fluorescens (B2, obtained from the culture collection of Department of Food

Science at Pennsylvania State University), Micrococcus luteus (ATCC 10240), Enterococcus faecalis (ATCC 51299), and Escherichia coli (ATCC 25922) was prepared as the initial inoculum to mimic the worst case scenario and increase the microbial population of the raw milk. Each bacterium was grown in 500 ml of Tryptic Soy Broth (TSB) for 24 h at their optimal temperatures. For P. fluorescens and M. luteus, the incubation temperature was set at 30°C, and 88 for E. faecalis and E. coli at 37°C. The population in each culture was found to be ~ 108 CFU/ml on average.

4.3.2 Description of the Pilot Milking System for One-step Cleaning Trial

A 114 L (30 gal) single bowl wash sink was used for the one-step pilot milking system

CIP (fig. 4-1). Other components that are essential to the milking system CIP are included in the pilot system as previously described (Dev et al., 2014). The test portion of the milking system pipeline consisted of a set of straight pipes with a total length of 24.4 m (80 ft) and eight 90° elbows. The CIP process of the pilot milking system is powered by vacuum, set at a 50 kPa constant level. A 38 L (10 gal) receiver jar was installed at the end of the milking system pipeline for temporary storage then either to redirect the cleaning solution back to the sink or to drain directly through a milk pump. The slope of the pipeline in the system was set to 0.8% to facilitate water draining. When conducting CIP, the solution fluid dynamic properties were adjusted to be at a mean slug velocity of 9.1 ± 1 m/s (30 ± 3 ft/s) and 3 slugs per min with an air admission rate of 0.12 m3/min (30 gal/min).

pH sensor (Model CSIM 11, Campbell Scientific, Inc., Logan, UT) and ORP sensor

(Model CSIM 11-ORP, Campbell Scientific, Inc., Logan, UT) along with several thermocouples

(PP-T-24, OMEGA Engineering, Inc., Stamford, CT) were used to monitor the pH, ORP and temperature during the CIP process. pH and ORP sensors were placed near the bottom of the wash sink. Two thermocouples were placed at each end close to the bottom of the wash sink; three placed at the outlets of the alkaline EO water, acidic EO water, and tap water faucets, and one inserted at the end of the return line of the pilot milking system to monitor the returning CIP solution temperature.

Figure 4-1. Pilot plant milking system and sampling location schematic of the pipes and elbows

(P1 – P8 represent pipe sampling locations and E1 – E8 represent elbow sampling

locations (revised from Dev et al., 2014; not to scale).

4.3.3 Preparation of Blended EO Water Solutions

EO water was generated with an EO water generator (Model ROX60SA) (Hoshizaki

Electric Co. Ltd., Sakae, Japan). Alkaline EO water was heated before use by a 305 L (80 gal) capacity tank water heater (Model RUE PRO-80-2, Ruud Manufacturing Co., Atlanta, GA) and acidic EO water was heated using a tankless heater (Model EX1608TC, Eemax Inc., Oxford, CT).

The blending process was carried out by adding heated acidic EO water to the heated alkaline EO water. In this manner, chlorine loss from the acidic EO water could be minimized and the final blended EO water temperature could be more controllable by adjusting the flow rate of the

90 tankless heater. Both EO water solutions were freshly generated before use and properties tested, both for heated and unheated solutions (table 4-1). For the unheated EO water solutions, the acidic EO water had pH of about 2.6, ORP of about 1150 mV, and free chlorine content of about

80 ppm; whereas the unheated alkaline EO water had pH of about 11.5 and ORP of about -850 mV. The chlorine content of the acidic EO water was measured by using a chlorine test kit (Hach,

Inc., Loveland, CO). When heated, the acidic EO water had pH of about 2.8, ORP of about 1150 mV, and free chlorine content around 65 ppm, which depends on the heated temperature, whereas the alkaline EO water had pH of about 10.5 and ORP of about -800 mV.

Table 4-1. EO water solution property comparisons.

Unheated solution Heated solution Acidic EO water Alkaline EO water Acidic EO water Alkaline EO water pH 2.6 11.5 2.8 10.5 ORP (mV) 1150 -850 1150 -800 Free chlorine 80 0 65 0 (ppm)

4.3.4 Preparation of the Milking System for CIP Process

4.3.4.1 Shock Cleaning

Shock cleaning of the pilot milking system was conducted before each experiment to create a consistent base line of the system. The term “shock cleaning” comes from a more powerful, namely more concentrated, cleaning solution for the milking system CIP:

i. Alkaline wash: 355 ml (12 oz) of Dairy Cycle 3 (Chemland Inc., Kansas City, MO) diluted into 68 L (18 gal) of alkaline wash solution with a starting temperature of 80°C and a circulation time of 10 min;

ii. Water rinse: 38 L (10 gal) of tap water heated to 40°C for rinsing the system without recirculation;

iii. Acid wash: 107 ml (3.6 oz) of Dairy M. S. R. 50 (Chemland Inc., Kansas City, MO) diluted into 68 L (18 gal) of acid wash solution with a starting temperature of 80°C and a circulation time of 10 min;

iv. Sanitizing rinse: right before each experiment, 27 ml (0.9 oz) of LCS (Classic

Technologies., Kansas City, MO) diluted into 68 L (18 gal) of sanitizing solution with a starting temperature of 40°C and no recirculation.

4.3.4.2 Soiling the System

Fresh raw milk was collected from Penn State Dairy Barn by directly milking the cow into a 38 L (10 gal) stainless steel milk can. The collected milk was reheated to about 40°C in the lab to compensate the heat loss during transportation. Each prepared culture was centrifuged at

4000 x g for 40 min (Model Sorvall T12, Thermo Fisher Scientific Inc., Waltham, MA), then the bacterial cells were re-suspended again using about 250 ml (0.05 gal) of raw milk then added back to the remaining 37.1 L (9.8 gal) of raw milk, which resulted in ~107 CFU/ml on average.

The inoculated milk was then used to soil the milking system. Under the vacuum, milk was drawn into the system pipeline and accumulated in the receiver jar then drained out without recirculation. The milk drawn into the system was in turbulent state with a Reynolds number to the order of magnitude of 105, which was calculated based on the average mean velocity of 9.1 ±

1 m/s as recommended by Dairy Practices Council for the characteristic dimension of 0.03683 m and the kinematic viscosity of milk is 1.13*10-6 m2/s at 20°C (Dev et al., 2014). The soiling process consisted of three repetitive sections; each one had a milk soiling and an air drying subsection. During the milk soiling subsection, approximately one third, 12.7 L (3.3 gal) of milk was introduced into the system pipeline under vacuum. After that, still with the vacuum on,

92 ambient air was introduced into the system for 10 min as the air drying subsection. The milk was not recirculated within the pilot milking system pipeline to minimize churning, cream separation, and other physio-chemical changes in the milk which would potentially affect the soiling process

(table 4-2).

Table 4-2. Operation process comparison between the blended EO water and commercial one-

step chemicals for pilot milking system CIP.

Operating cycle

Operating parameter Warm Blended EO Commercial Sanitize Soiling water water one-step rinse cleaning cleaning

Starting temperature 40 38 40 Variable 70 (°C)

Quantity (L) 68 38 (12.7×3 parts) 38 57 57

Air injection No No No Yes Yes

Recirculation No No No Yes Yes

32 (<1 min/part Operation time (min) 3.5 and 10 min 2 Variable 10 break/part) Note: shock cleaning of the pilot milking system was conducted before each experiment to create a consistent base line of the system.

4.3.5 Experimental Design

Three factors were included in this study, namely the cleaning time (10, 15 and 20 min), the starting temperature (50°C, 60°C and 70°C) of the blended EO water solution, and the acidic

EO water percentage (25%, 42.5%, and 60%) in the blended EO water solution. Preliminary experiments were conducted to determine the acidic EO water percentage range in the blended

EO water solution by testing the CIP performance for all sampling locations using different acidic

EO water percentages (0% to 100%) with a fixed mid-point for the other two factors’ (the cleaning time and the starting temperature of the blended EO water solution) respective ranges.

Table 4-3. Three factor Box-Behnken experimental design for CIP optimization process.

Run Order Time (min) Temperature (°C) Acid percentage

1 10 70 42.5 2 15 70 60 3 20 60 25 4 15 50 60 5 15 60 42.5 6 20 70 42.5 7 20 60 60 8 15 60 42.5 9 10 50 42.5 10 10 60 25 11 20 50 42.5 12 15 70 25 13 10 60 60 14 15 50 25 15 15 60 42.5

Box-Behnken three-factor response surface was used to determine the optimal of these three parameters (table 4-3). In addition, the total quantity of the blended EO water solution was fixed at 57 L (15 gal) based on the calculations of the length of the system pipeline and other milking system components (Reinemann, 1995). When the optimal condition of each factor was achieved, three validation runs were conducted at this optimal condition to verify the optimization result.

Two commercial one-step CIP chemicals were used for CIP effectiveness and economical cost comparisons. They were obtained from different manufactures and denoted as commercial product 1 (CP1) and commercial product 2 (CP2). The labels of both products showed “contains sulfuric acid, methane sulfonic acid and hydrogen peroxide”. CIPs using these two chemicals were performed based on the label instructions, which were: a 10 min cleaning time, a 70°C starting cleaning solution temperature and a solution concentration of 29.6 ml (1 oz) chemical per 3.79 L (1 gal) water. For each of the commercial one-step CIP, three replications were conducted.

4.3.6 Evaluation

Sampling locations are categorized into pipes, elbows, and other milking system components which include gaskets, liners, and milk hose (fig. 4-2). Eight straight pipe sampling locations and eight elbow sampling locations were selected along the milking system pipeline.

The swabbing areas of sampling locations of pipes and elbows followed previously described procedures (Dev et al., 2014). For each sampling location, two types of evaluation methods were used: Adenosine Triphosphate (ATP) bioluminescence for the presence of soils and sterile calcium alginate-tipped applicators swabbing for microbial cells presence. Samples were collected from each sampling location by disassembling the connections (also shown in figure 4-

2) and swabbing the inner surfaces of elbows and pipes (Dev et al., 2014). With reference to the cleaning solution flow direction, the right half of the connections of sampling locations were sampled using ATP swabs for bioluminescence analyses and the left half using sterile calcium alginate-tipped applicators for analyzing the bacterial presence. Other milking system components of gaskets, liners, and milk hose were swabbed for ATP bioluminescence and bacterial presence as well, but the only difference is the sample was taken by swabbing over the entire inner circle area instead of half inner circle area due to the limited availability of milk contact surfaces for swabbing in these milking system components.

Figure 4-2. Location sampling protocols of the pipes, elbows and other milking system

components.

4.3.7 Analysis

Inner surfaces of all sampling locations were evaluated using ATP bioluminescence

PocketSwab Plus swabs (Charm Science, Inc., Lawrence, MA) for soils and sterile calcium alginate-tipped applicators (Puritan Medical Production Co. LLC, Guilford, ME) for analyses of bacterial cell presence.

4.3.7.1 ATP Bioluminescence

The ATP swabs’ firefly enzyme luciferase gave a quantitative measurement in terms of

Relative Light Unit (RLU) by using a novaLUM palm-sized luminometer (Charm Science, Inc.,

Lawrence, MA); the RLU readings served as an indicator of the surface cleanliness. An RLU reading of “0” represents the surface was clean; otherwise, the higher the RLU readings showed, the “dirtier” surfaces were. For further discussion purposes, RLU reduction percentage and RLU log reduction were used for CIP performance comparison. The RLU reduction percentage is the after CIP RLU reading compared to the before CIP RLU reading; and higher RLU reduction percentage represented a more effective CIP process. The RLU log reduction is difference between the natural logarithm of before CIP RLU reading and the natural logarithm of after CIP

RLU reading. The equations used for the calculations of RLU reduction percentage and RLU log reduction are presented as follows:

RLU Reduction Percentage = (1 – RLU after CIP / RLU before CIP) * 100% (1)

RLU Log Reduction = log(RLU before CIP + 1) – log(RLU after CIP + 1) (2)

In equation 2, the numeral 1 was added to avoid indeterminacy of log when the measured

RLU is 0. Since the RLU values measured in this study were mostly in the order of 106 or higher, 97 the error introduced is negligible. By definition of the RLU reduction percentage, a 100% RLU reduction represents an after CIP RLU reading of 0.

4.3.7.2 Calcium Alginate Sampling

Bacterial samples were collected with sterile calcium alginate-tipped applicators (Puritan

Medical Production Co. LLC, Guilford, ME), which were placed in a TSB medium, and then incubated at 37°C for 48 h for possible microbial growth. The clear medium observed visually after incubation was accepted as “negative” and the opaque (turbid) ones as “positive.” For further discussion purposes, negative enrichment percentage was calculated by percentage of the number of negative samples (referred as “negative”) with respect to the total number of samples

(number of negative samples plus positive samples) (referred as “total”) for all three sampling categories (eq. 3). By this definition, higher negative enrichment percentage represents a more sanitized surface condition.

Negative Enrichment Percentage = (negative/ total) * 100% (3)

4.3.7.3 Operational Cost Analysis

The operational cost of one CIP process was calculated based on the following analyses.

Material cost and energy consumption during the cleaning cycle were taken into consideration.

These included: (1) the salt to generate the EO water solutions and the salt to soften supply water which reduces the supply water hardness to facilitate EO water generation, (2) commercial one- step chemicals, and (3) electricity usage which includes the solution heating and mechanical consumption for each cycle of one CIP process. The power usage of the vacuum pump and the milk pump was measured using an ELITEproTM recording poly phase power meter (Dent

Instruments, Bend, OR) to acquire more accurate measurements during the appliance working state. The general comparison of the operational cost is listed in table 4-6.

4.3.7.4 Statistical Analysis

Response surface experimental table was generated and results analyzed using Minitab

(Version 16.2, Minitab Inc, State College, PA). Three replications were conducted at the optimal blended EO water conditions. The significant differences in mean values of the RLU reduction percentages and negative enrichment percentages of sampling locations of pipes, elbows and other milking system components were determined using Tukey’s method at the 95% confidence interval.

4.4 Results and Discussions

This study investigated the use of the blended EO water solution as an alternative for a pilot milking system one-step CIP. An optimization process was conducted firstly to determine the optimal condition of three affecting parameters: the cleaning time, the starting temperature of the blended EO water solution and the acidic EO water percentage in the blended EO water solution using a Box-Behnken three-factor response surface method. The CIP performance was compared using the optimized blended EO water solution and two commercial one-step cleaning chemicals; and the evaluation methods included the ATP bioluminescence and the bacterial presence on the sampling locations of pipes, elbows and other milking system components. The operational cost for the blended EO water solution at its optimal condition was compared with the commercial one-step CIPs.

4.4.1 Determining the Range of Parameters Evaluated for One-step CIP

The cleaning time of the blended EO water solution was evaluated in the range from 10 to 20 min. The recommended cleaning time was to hold “a minimum of 20 slugs for each CIP process” (DPC, 2010), and based on the configuration of the pilot milking system, a 10 min wash circulation for one cycle was needed (Dev et al., 2014). Therefore, the cleaning time of the blended EO water solution varies from one complete CIP cycle of 10 min (one-step CIP) to two complete CIP cycles of 20 min (conventional alkaline wash followed by acid wash CIP). The starting temperature of the blended EO water solution was set between 50°C to 70°C, based on the following reasons: i) The previously optimized alkaline solution temperature was 70°C

(highest) and acid solution was 45°C (lowest) for the pilot milking system, which provided a rough range of the cleaning solution starting temperature, and ii) The one-step CIP needs to remove both organic (mostly by alkaline wash cycle conventionally) and mineral materials

(mostly by acid wash cycle conventionally) from the milk-contact surfaces. Therefore, 50°C to

70°C temperature range was selected to find out the optimal temperature for the combined steps.

There were no previous published studies on the range of the acidic EO water percentage in the blended EO water solution for the milking system one-step cleaning, and neither were any published recommendations available. Therefore, an acidic EO water percentage in the blended

EO water solution ranging from 0% to 100% was evaluated for the pilot milking system CIP during preliminary experiments. By keeping the cleaning time and the starting temperature of the blended EO water solution at mid-point of their respective ranges, namely at 15 min for cleaning time and 60°C for the blended EO water solution starting temperature, preliminary experiments were conducted to determine the acidic EO water percentage range (fig. 4-3 and 4-4).

100

100.00%

99.90%

99.80% RLU reduction Percentage pipes 99.70% elbows others

99.60%

99.50%

Acidic EO water percentage

Figure 4-3. RLU reduction percentages at different acidic EO water percentages for sampling

categories of pipes, elbows and other milking system components of gaskets, liners, and

milk hose.

101

100%

90%

80%

70%

60% Negative Enrichment 50% Percentage pipelines 40% elbows 30% others

20%

10%

Acidic EO water percentage

Figure 4-4. Negative enrichment percentages at different acidic EO water percentages for

sampling categories of pipes, elbows and other milking system components of gaskets,

liners, and milk hose.

Results showed that the CIP was not effective, when the blended EO water solution approaches near neutral pH range; for example, when the acidic EO water percentage was

40.00% and the corresponding blended EO water pH was 6.2, the average RLU reduction percentage was 99.83%, ranking second lowest of all trials and the average bacterial enrichment was 23.7%, ranking the lowest of all trials. The more acidic solutions, on the other hand, had relative better CIP performance compared to the more alkaline solutions (fig. 4-3 and 4-4). RLU reduction percentage and negative enrichment percentage were used for the analyses. When the acidic EO water percentage reached and above 46.7%, the RLU reduction percentages for

102 sampling locations of pipes and elbows reached 100%; and the negative enrichment percentages for pipes and elbows reached higher than 60%. On the other hand, when the acidic EO water percentage dropped to 26.7% and below, the RLU reduction percentages for sampling locations of pipes and elbows were above 99.9% until when using a pure alkaline EO water solution dropped to around 99.8% (fig. 4-3). The negative enrichment percentages of the more alkaline blended EO water solutions were relative higher compared to near neutral conditions; but not as high as the more acidic blended EO water solutions (fig. 4-4).

Upon review, it was determined that the more acidic blended EO water solutions did result in a better CIP performance when holding the other two factors, i.e., the cleaning time and the blended EO water solution starting temperature, as constant (fig. 4-3 and 4-4). From the results, it was observed that there is no significant increase in the RLU reduction percentage, when the acidic EO water percentage is higher than 46.70%. Therefore, a middle range of acidic

EO water percentage was chosen for the optimization process. A more important reason to choose more acidic solution over more alkaline solution is that, based on the after CIP RLU readings from the preliminary experiments and a set of cut-off values presented in previous studies of system materials of rubber (cut-off RLU of 4,500) and stainless steel (cut-off RLU of 1,000)

(Wang et al., 2013), the more acidic CIP trials resulted in a below the cut-off value RLU reading, while the more alkaline CIP trial could not. Therefore, it was determined that the acidic EO water percentage for the response surface method should be from 25% to 60%.

4.4.2 Optimization of the Blended EO Water CIP Process

The Box-Behnken three-factor response surface experimental design is shown in table 4-

3. A total of 15 runs were conducted to get the optimal condition of each factor. Table 4-4 shows the results of the response surface design, including the result of RLU reduction percentages, the

103 result of RLU log reductions and the result of negative enrichment percentages, for the sampling locations of pipes, elbows and other milking system components of gaskets, liners, and milk hose.

Run # 5, 8 and 15 are experiments at the center point of each factor, which showed a relative consistency for RLU reduction percentage (table 4-4).

Table 4-4. Response surface method result for each sampling category.

Run RLU Reduction Negative Enrichment RLU Log Reduction Order Percentage Percentage

Pipe Elbow Other Pipe Elbow Other Pipe Elbow Other

1 99.86 99.88 99.49 2.88 2.94 2.44 25 25 33

2 99.99 100 99.71 3.84 4.36 2.77 63 50 33

3 99.89 99.91 97.99 2.95 3.07 1.70 25 25 33

4 99.99 99.99 99.90 4.24 4.16 3.12 75 63 33

5 99.99 99.97 99.72 4.00 3.56 2.45 50 38 33

6 99.98 99.99 99.81 3.79 3.98 2.71 50 63 33

7 100 100 100 6.34 6.34 6.25 75 75 68

8 99.97 99.99 99.69 3.59 3.92 2.69 38 50 33

9 99.84 99.84 99.84 2.80 2.81 2.87 13 25 33

10 99.70 99.73 99.45 2.54 2.61 2.35 13 13 33

11 99.99 100 99.91 4.11 4.47 3.03 63 75 33

12 99.89 99.88 98.54 2.97 2.91 1.93 25 25 33

13 99.98 100 99.86 3.73 6.54 2.88 75 50 33

14 99.77 99.80 99.38 2.68 2.71 2.30 13 25 33

15 99.98 99.97 99.49 3.68 3.53 2.79 50 38 33

104

The result of RLU log reduction demonstrated how much ATP was removed for each experiment in the absolute value. The RLU log removal was different for each experiment, which resulted from i) the CIP performance being different for each blended EO water solution, and ii) the initial RLU reading differed due to the initial quality of the raw milk acquired on the experimental day. Any conclusions solely based on the RLU log reduction are not sufficiently reliable; therefore, the term RLU reduction percentage, which is a relative number, was used for the remainder of the analyses in this study. However, despite the differences caused by the raw milk quality for every experiment, the CIP process was able to remove at least 2.5 log of residue

ATP on the surfaces.

When using RLU reduction percentage as a response for the optimization process, the models for the sampling locations of pipes, elbows and other milking system components were suggested by Minitab (eqs. 4-6), in which T stands for time in minutes, AP stands for acid percentage (%), and TEMP stands for temperature in °C:

Pipe RLU Reduction Percentage (%) = 99.980 + 0.060 * T + 0.016 * TEMP + 0.089 * AP –

0.040 * T2 – 0.023 * TEMP2 – 0.048 * AP2 – 0.008 * T * TEMP – 0.043 * AP * T – 0.030 * AP *

TEMP, R2 = 97.72% (4)

Elbow RLU Reduction Percentage (%) = 99.977 + 0.056 * T + 0.015 * TEMP + 0.084 * AP –

0.028 * T2 – 0.021 * TEMP2 – 0.038 * AP2 – 0.013 * T * TEMP – 0.045 * AP * T – 0.018 * AP *

TEMP, R2 = 98.04% (5)

Other RLU Reduction Percentage (%) = 99.670 – 0.116 * T – 0.019 * TEMP + 0.514 * AP +

0.018 * T2 + 0.075 * TEMP2 – 0.363 * AP2 – 0.063 * T * TEMP – 0.400 * AP * T – 0.163 * AP *

TEMP, R2 = 88.70% (6)

105

The 3D surface contour plots of sampling categories of pipes and elbows are shown in figure 4-5 and figure 4-6. From the trend of the surface plots and also the statistical analyses, the cleaning time and acidic EO water percentage of the blended EO water solution significantly affected the RLU reduction percentage for sampling locations of pipes and elbows (P<0.05) – higher starting temperature and longer cleaning time are more favorable to achieve a more satisfactory CIP performance. For example, the RLU reduction percentage of elbow with a shorter time (10 min) and a lower acidic EO water percentage (25%), and a median temperature of 60°C was about 99.7%, but it went up to almost 100% if the cleaning time is longer (20 min) with a higher acidic EO water percentage (60%) without changing the temperature setting (fig. 4-

6). The starting temperature of the blended EO water solution generated mixed results – the increased temperature would increase the reaction rate thus enhancing the cleaning performance, but the increased temperature reduced the amount of chlorine in the blended EO water solution, which resulted in a reduced sanitizing power. From the statistical analysis, the temperature effect was not significant for all three sampling locations of pipes (P=0.109), elbows (P=0.090) and other milking system components (P=0.153).

When comparing the results for sampling locations of pipes, elbows and other milking system components (table 4-4), it was observed that the stainless steel materials such as pipes and elbows are more easily cleaned compared to other milking system components made from materials such as rubber and polyvinyl chloride (PVC); the average RLU reduction percentage of sampling location of pipes was as high as 99.92% and 99.93% for elbows, while for the other milking system components of gaskets, liners, and milk hose, the average RLU reduction percentage was only about 99.52%.

106

Figure 4-5. 3D surface plots of RLU reduction percentage for sampling locations of pipes.

Figure 4-6. 3D surface plots of RLU reduction percentage for sampling locations of elbows.

107

Figure 4-7. CIP performance comparison among two commercial one-step cleaning chemicals

and blended EO water at its optimal conditions with CP1 represents commercial product

1 and CP2 for commercial product 2: (a) RLU reduction percentage comparison; (b) RLU

log reduction comparison; (c) negative enrichment percentage comparison.

Rubber has been shown to have “caverns” and crevices over the surfaces. A scanning electron microscopy analysis of a rubber liner sample surface collected directly from an operating dairy farm, results showed the presence of cracks on the surface (Latorre et al., 2010). One study demonstrated a greater number of Pseudomonas fragi, Listeria monocytogenes and Bacillus cereus attachment onto buna-N-rubber and teflon (Gaspar-Rolle, 1991). The highly developed porosity on the rubber surfaces makes CIP performance harder to achieve compared to smoother stainless steel surfaces. It is recommended that the rubber goods should be inspected and replaced

108 on a regular basis on a dairy farm, twice a year (recommended), to prevent the rubber from aging and becoming more porous (DPC, 2010).

By running the Minitab response optimizer, it was suggested that a cleaning time of 17 min, a starting temperature of 59°C and an acidic EO water percentage of 60% were the optimal conditions for the blended EO water solution CIP. Three validation runs were conducted at this optimal condition and results showed good reproducibility (fig. 4-7). At this condition, 100%

RLU reduction percentages were achieved for both sampling locations of pipes and elbows experimentally, as predicted from Minitab desirability calculation. The other milking system components including the gaskets, liners, and milk hose resulted in a 99.95% RLU reduction percentage experimentally, which is much higher compared to the average of 99.52% during the

15 runs of the optimization process. The negative enrichment percentages of sampling locations of pipes and elbows reached around 80% on average, and as stated above of the material property differences, no improvement in negative enrichment percentage was observed (33.3%) for the other milking system components including gaskets, liners, and milk hose.

4.4.3 CIP Performance Comparison between the Optimal Blended EO Water and the

Commercial One-step Chemicals

Two types of commercial one-step CIP chemicals were used to conduct the CIP performance comparisons; they were from different companies and hereby denoted as commercial product 1 (CP1) and commercial product 2 (CP2). Both of them have a blend of organic and inorganic acids and a claimed good wetting capability and effective removal capability of lipid, protein and mineral within one CIP cycle. Based on the instructions of the product labels, parameters for the commercial one-step CIP were set at a cleaning time of 10 min, a starting cleaning solution temperature of 70°C, and a solution concentration of 29.6 ml (1 oz)

109 chemical per 3.79 L (1 gal) water. For the blended EO water solution, the optimal condition (the cleaning time of 17 min, starting temperature of 59°C and an acidic EO water percentage of 60% in the blended EO water solution) was used for comparison (table 4-5).

Table 4-5. CIP parameter comparisons between using two different commercial one-step

chemicals and the blended EO water at its optimal condition (CP1 represents commercial

product 1 and CP2 for commercial product 2).

CP1 & CP2 Optimized blended EO water

Solution quantity (L) 57 57

Cleaning duration (min) 10 17

Starting temperature (°C) 70 59

29.6 ml (1 oz) chemical / Concentration 60% acidic EO water 3.79 L (1 gal) water

The results of the CIPs were compared in figure 4-7. The RLU reduction percentage was not significantly different among the two commercial one-step products and the optimal blended

EO water solution (p>0.05). The average RLU reduction percentage of optimal blended EO water was 100% while the two commercial one-step chemicals were 99.96% and 99.97%, respectively, for sampling locations of pipes; the average RLU reduction percentage of optimal blended EO water was 100% while the two commercial one-step chemicals were 100% and 99.95%, respectively, for sampling locations for elbows (fig. 4-7a). The RLU reduction percentage is a relative number compared to the initial RLU reading from the soiling process; therefore, given the high initial soiling RLU readings (on the order of magnitude of 106 or higher), it is not easy to differentiate from the RLU reduction percentage among different CIP methods. However, when

110 comparing the bacterial presence data (fig. 4-7c), the difference between the blended EO water and the two commercial products are noticeable. For the negative enrichment percentage, blended

EO water performed better compared to the commercial one-step chemicals for the stainless steel surfaces including sampling locations of pipes and elbows on average. The comparison among the commercial CIPs and optimal blended EO water CIP was not statistically different (p>0.05), however, the average negative enrichment percentage was 83.3% for the optimal blended EO water CIP but only 50.0% and 54.2% for the two commercial one-step CIPs, respectively (fig. 4-

7c). This is not surprising given the past studies of using blended EO water solution as a disinfectant for produce and food processing plants. Other milking system components, including gaskets, liners, and milk hose, however, did not have a significant difference (p>0.05) among CIP methods (fig. 4-7). The natural surface condition of these milking system components made them hard to be cleaned sufficiently and the crevices provided the bacteria a habitat for growth; therefore, the non-significant differences in results were not surprising. Generally, all of these

CIP methods demonstrated satisfactory CIP performance results, indicating that they are capable of removing the organic and inorganic soils as well as reducing the bacteria load and keeping the milk contact surfaces clean and sanitized without bringing in potential hazard.

There have been studies that explain how the blended EO water solution worked as a disinfectant, but not as a cleaning agent. Most of the studies demonstrated the effect of both chlorine and ORP on killing the microorganisms – high ORP of the solution disrupts the outer cell membrane, thus facilitating the transfer of HOCl across the cell membrane, resulting in further oxidation of intracellular reactions and respiratory pathways such as the GSSH/2GSH cellular redox couple (Liao et al., 2007). Others attributed the fungicidal efficiencies of near neutral EO water to the •OH radical (Xiong et al., 2010). For the cleaning process mechanism using blended EO water solution, on the other hand, none is found in the current literature review; the “wetting power” which is a necessity of detergent was not extensively studied for the blended

111

EO water solution. More studies are needed to clearly indicate the mechanism of the blended EO water functions as a cleaning agent.

4.4.4 Operational Cost Comparison between the Optimal Blended EO Water CIP and the

Commercial One-step CIP

Cost analyses were performed between using the blended EO water at its optimal condition and the commercial one-step chemicals at their suggested CIP conditions as described above. Here only the operational cost was analyzed, i.e., the purchase of the EO water generator, the maintenance and labor cost was excluded in the operational cost calculation. Table 4-6 shows the operational cost comparisons.

The total cost was divided into three major categories, namely the cleaning agent cost, the heating usage cost and the mechanical usage cost. For the cleaning agent cost, the EO water generation uses only table salt, which is easily accessible and less expensive; the commercial one- step cleaning agent on the other hand, is much more expensive. The difference in the heating usage cost between these two methods was that when using the commercial one-step CIP, a starting temperature of the cleaning solution of 70°C was required, whereas the blended EO water

CIP at its optimal condition only needed a starting temperature of 59°C. The mechanical usage for the two methods was slightly different, because for the blended EO water CIP, the mechanical usage included the operation of the external water softener, the generation of EO water solutions, and the operations of the vacuum pump and milk pump; and for the commercial one-step CIP, only the operations of the vacuum pump and milk pump were included in the mechanical usage, since the water hardness was not required for the commercial one-step chemicals.

Overall, the operational cost for the blended EO water CIP at its optimal condition was calculated as $0.55, whereas the operational cost for the commercial one-step CIP was calculated 112 as $2.82 at the suggested concentration by the manufacturers (table 4-6). It is clear to see that using the blended EO water at its optimal condition for the pilot milking system CIP had an advantage over using commercial one-step CIP: it is about 80% less expensive than the commercial one-step CIP. Even when using the lowest concentration required for the commercial one-step CIP the cost of using blended EO water CIP could still be 2/3 less expensive than using the commercial one-step CIP for the pilot milking system.

In addition to the reduced cost of the cleaning agent, the heating cost comparison demonstrated that using the blended EO water at its optimal condition is also a potential energy efficient approach. The heating usage difference between the blended EO water CIP and the commercial one-step CIP might not seem to be significantly different in this study due to the lab scale set-up of only 24.4 m (80 ft) in length and 57 L (15 gal) in the amount of cleaning solution.

However, an example of a previous study (Wang et al., 2013), the cleaning solution quantity needed was 114 L (30 gal) on a mid-sized commercial dairy farm. In this case, the heating usage gap between using the blended EO water and the commercial one-step chemicals is larger – it is calculated to be $0.14 difference for one CIP process. The cleaning time of the blended EO water

CIP (17 min) is longer than the commercial one-step CIP (10 min), but from the cost comparison

(table 4-6), the mechanical usage and heating cost are almost identical ($0.49 for blended EO water and $0.48 for commercial one-step); it is the large differential of cost of chemicals both in absolute ($2.34-$0.06=$2.28) and relative (100×$2.34/$0.06=3900%) terms that makes the blended EO water ($0.06) cost effective compared with commercial one-step ($2.34) CIP.

Therefore, the relative longer cleaning time for the blended EO water CIP should still maintain the cost advantage even if the milking system is scaled up to a larger size.

However, despite the promising results of using the blended EO water as an alternative for the pilot milking system one-step CIP, both technically and economically, the capital cost of the EO water generator must be further evaluated.

113

Table 4-6. Operational cost comparison between the blended EO water and commercial one-step

CIP on the pilot milking system.

Blended EO Water CIP Commercial one-step CIP

Cost ($) Percentage Cost ($) Percentage

Cleaning agent 0.06 10.9 2.34 83.0 Heating usage 0.33 60.0 0.40 14.2 Mechanical usage 0.16 29.1 0.08 2.8

Total 0.55 100 2.82 100 Essential equations used for calculations:

Q = Cp×m×Δt (Q: heat; Cp: Specific heat; m: mass; Δt: temperature increment)

W = V×A (W: Power; V: Voltage; A: Amperage)

4.5 Conclusions

In conclusion, an optimal condition of the blended EO water solution for the pilot milking system CIP is achieved in this study. The combination consists of a cleaning time of 17 min, a starting temperature of the blended EO water solution of 59°C, and an acidic EO water percentage of 60% in the blended EO water solution. Three validations at this optimal condition showed 100% RLU reduction percentages for both sampling locations of pipes and elbows experimentally, which matched the model estimation of CIP performance. When comparing the

CIP performance of the optimal blended EO water with two commercial one-step chemicals, the

RLU reduction percentage results showed that the optimal blended EO water was doing as good as the two commercial one-step CIP chemicals. In addition, the blended EO water resulted in a higher negative enrichment percentage on average compared to the commercial one-step CIP 114 chemicals, indicating the blended EO water at its optimal condition had better disinfecting capability. Moreover, the cost comparisons showed that the operational cost for the blended EO water CIP at its optimal condition was about 80% less expensive than using the commercial one- step chemicals for the pilot milking system CIP. Based on this research, it is concluded that the blended EO water at its optimal condition has the potential to be adapted as an alternative one- step CIP method for milking system, and possibly other food processing equipment in terms of both effectiveness and economics.

4.6 Acknowledgements

Funding for this project was provided in-part by a USDA Special Research Grant (No.

Roderick Thomas, Randall Bock, and all Penn State Dairy Barn personnel for their help in the project.

4.7 References

Dev, S. R. S., Demicri, A., Graves, R. E., & Puri, V. M. (2014). Optimization and modeling of an

electrolyzed oxidizing water based Clean-In-Place technique for farm milking systems

using a pilot-scale milking system. J. Food Engr., 135, 1 – 10.

DPC. (2010). Guidelines for Installation, Cleaning, and Sanitizing of Large and Multiple

Receiver Parlor Milking Systems. The Dairy Practices Council Cuidelines (pp. 46).

Richboro, PA.: The Dairy Practices Council.

115

FAOSTAT. (2012). Total production of cow milk, whole fresh cow. Retrieved from

http://faostat3.fao.org/faostat-gateway/go/to/browse/Q/QL/E.

FDA. (2012). The Dangers of Raw Milk: Unpasteurized Milk Can Pose a Serious Health Risk.

Retrieved from

http://www.fda.gov/downloads/Food/FoodborneIllnessContaminants/UCM239493.pdf.

Gaspar-Rolle, M. N. P. (1991). Attachment of bacteria to teflon and buna-n-rubber gasket

materials. Virginia Polytechnic Institute and State University. Retrieved from

http://vtechworks.lib.vt.edu/handle/10919/39818,

Guentzel, J. L., Callan, M. A., Liang Lam, K., Emmons, S. A., & Dunham, V. L. (2011).

Evaluation of electrolyzed oxidizing water for phytotoxic effects and pre-harvest

management of gray mold disease on strawberry plants. Crop Prot., 30(10), 1274–1279.

doi:10.1016/j.cropro.2011.05.021

Guentzel, J. L., Lam, K. L., Callan, M. A., Emmons, S. A., & Dunham, V. L. (2010). Postharvest

management of gray mold and brown rot on surfaces of peaches and grapes using

electrolyzed oxidizing water. Int. J. Food Microbiol., 143(1-2), 54–60.

doi:10.1016/j.ijfoodmicro.2010.07.028

Guentzel, J. L., Liang Lam, K., Callan, M. A., Emmons, S. A., & Dunham, V. L. (2008).

Reduction of bacteria on spinach, lettuce, and surfaces in food service areas using neutral

electrolyzed oxidizing water. Food Microbiol., 25(1), 36–41.

Jones, G.M. (2009). Cleaning and saniziting milking equipment. Retrieved from

http://pubs.ext.vt.edu/404/404-400/404-400.html.

Latorre, A. A., Van Kessel, J. S., Karns, J. S., Zurakowski, M. J., Pradhan, A. K., Boor, K. J.,

Jayarao, B. M., Houser, B. A., Daugherty, C. S., & Schukken, Y. H. (2010). Biofilm in

milking equipment on a dairy farm as a potential source of bulk tank milk contamination

with Listeria monocytogenes. J. Dairy Sci., 93(6), 2792–802. doi:10.3168/jds.2009-2717

116

Liao, L. B., Chen, W. M., & Xiao, X. M. (2007). The generation and inactivation mechanism of

oxidation–reduction potential of electrolyzed oxidizing water. J. Food Engr., 78(4),

1326–1332. doi:10.1016/j.jfoodeng.2006.01.004

Parr, K. (2013). New Exacta One-Step CIP Detergent. AgroChem Farm & Dairy Products.

Retrieved from

http://www.agrocheminc.com/images/brochures/exacta_press_release.pdf.

Reinemann, D. (1995). System design and performance testing for cleaning milking systems.

Proc. Designing a Modern Milking Center: Parlors, Milking Systems, Management and

Economics. Retrieved from

http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:System+Design+and+

Performance+Testing+for+Cleaning+Milking#0.

Walker, S., Demirci, A., Graves, R., Spencer, S., & Roberts, R. (2005a). Cleaning milking

systems using electrolyzed oxidizing water. Trans. ASAE, 01(48), 1827–1833.

Walker, S., Demirci, A., Graves, R., Spencer, S., & Roberts, R. (2005b). Response surface

modelling for cleaning and disinfecting materials used in milking systems with

electrolysed oxidizing water. Int. J. Dairy Tech., 58(2), 65–73.

Wang, X., Dev, S. R. S., Demirci, A., Graves, R. E., & Puri, V. M. (2013). Electrolyzed

Oxidizing Water for Cleaning-In-Place of On-Farm Milking Systems Performance

Evaluation and Assessment. Appl. Engr. Agrc., 29(5), 717–726.

Xiong, K., Liu, H.-J., Liu, R., & Li, L.-T. (2010). Differences in fungicidal efficiency against

Aspergillus flavus for neutralized and acidic electrolyzed oxidizing waters. Int. J. Food

Microbiol., 137(1), 67–75. doi:10.1016/j.ijfoodmicro.2009.10.032

117

CHAPTER 5

MATHEMATICAL MODELING AND CYCLE TIME REDUCTION OF

DEPOSIT REMOVAL FROM STAINLESS STEEL PIPELINE DURING

CLEANING-IN-PLACE OF MILKING SYSTEM WITH ELECTROLYZED

OXIDIZING WATER

5.1 Abstract

The safety of raw milk largely depends on using a clean milking system during the milk production. The milking system cleaning process widely used on dairy farms is a highly automated process called cleaning-in-place (CIP), which comprises of four cycles: i) warm water rinse; ii) alkaline wash; iii) acid wash; and iv) sanitizing rinse before the next milking event.

Electrolyzed oxidizing (EO) water is an emerging technology, which consists of acidic and alkaline solutions by the electrodialysis of dilute sodium chloride solution. Previous studies in our lab had shown that EO water can be an alternative for milking system CIP. Despite the progress made to enhance the CIP performance and evaluate alternative CIPs, the mechanisms behind the cleaning processes were still largely unclear. Therefore, this study was undertaken to evaluate the deposit removal rate during the EO water CIP process using a stainless steel surface evaluation simulator. Deposit removal data from the simulator formed the basis for developing mathematical models to describe the deposit removal process during the CIP process with EO water. Stainless steel straight pipe specimens were placed at the end of undisturbed entrance length along the simulator pipeline. The mass of the milk deposits on the inner surfaces of the specimens were measured using a high precision balance after the initial soiling, and then after certain time durations within the warm water rinse cycle, alkaline wash cycle, and acid wash cycle. A unified first order deposit removal rate model dependent on nth power of remaining deposit mass was 118 used for all three cycles. ATP bioluminescence method was also used as a validation approach at the end of each CIP cycle. Experimental results showed that the milk deposit on the inner surfaces of the specimens was removed rapidly by the warm water rinse within 10 s of rinse time. For the alkaline and acid wash cycles, the co-existence of a fast deposit removal at the beginning of the wash cycle and a slow deposit removal throughout the entire wash cycle was inferred. The proposed models matched the experimental data with small root mean square errors (0.23 mg/mg/m2 for the upstream locations and 0.07 mg/mg/m2 for the downstream locations) and satisfactory percent error differences (3.67% for the upstream locations and 0.93% for the downstream locations). Based on the experimental data and the proposed models, the time duration of the CIP process was shortened by 55% to 10 s warm water rinse, 3 min alkaline wash and 6 min acid wash was validated, which yielded an average deposit of 0.28 mg/mg/m2 at the end of the CIP as compared that of 0.29 mg/mg/m2 at the end of the original CIP, to achieve a satisfactory CIP performance for the simulator.

5.2 Introduction

The cleaning and sanitizing of milk processing pipelines are done using a highly automated cleaning-in-place (CIP) process to ensure the safety and quality of consumed milk.

The CIP process does not require disassembly and reassembly of the pipelines and other fittings of the system and, therefore, is widely adopted in food processing plants due to its high automation. Conventional CIP for the milking system on a dairy farm, usually performed immediately after the milking event, consists of four cycles: i) warm water rinse; ii) alkaline wash; iii) acid wash; and iv) a sanitizing rinse just before the next milking event. The warm water rinse removes the milk residual on the inner surfaces, and the alkaline wash removes organic materials such as lipids and proteins; and the acid wash removes the mineral deposits and

119 maintains the system in an acidified state to retard the growth of microorganisms. The sanitizing circulation, just prior to the next milking event, ensures the sanitization of the milking system.

Electrolyzed oxidizing water is a novel technology, which provided alkaline and acidic solutions (waters) generated through the electrodialysis of dilute sodium hydroxide solution. The acidic EO water has the potential to reach a pH as low as 2.6 and an oxidation-reduction potential

(ORP) as high as 1,150 mV. The alkaline EO water possesses a pH as high as 11.4 and a negative

ORP to -795 mV. Depending on needs, different properties of EO water can be generated by adjusting amperage and voltage. These properties of alkaline and acidic EO water fit the requirement of the milking system CIP (Hsu, 2005). Previous studies in our lab have demonstrated the efficacy of using EO water for the milking system CIP, both on a lab scale pilot milking system and on a commercial dairy farm (Dev et al., 2014, Wang et al., 2013). EO water has the advantage of having low operational cost and being environmentally benign.

There are several factors affecting the CIP performance including the type and strength of the cleaning chemical, the quantity and temperature of the solutions, and the fluid dynamics in the milking system pipelines. Recommendations have been established to provide guidance for these criteria (DPC, 2010). The properties of the cleaning chemical include pH, alkalinity, and chlorine concentration; these requirements vary based on the hardness of local water supply. The quantity and temperature of the cleaning solution can be calculated based on the pipeline length and farm environmental condition (Reinemann, 1995). The fluid dynamics of the cleaning solution can be adjusted based on the shear force needed to sufficiently remove the soils on the milk contact surfaces. Most of the recommendations greatly emphasize the importance of the first CIP cycle, warm water rinse, asserting the removal of a large amount of deposits if conducted per the recommendations. However, the quantitative contribution of this warm water rinse cycle as well as the contributions of rest of the CIP cycles of alkaline and acid washes to the CIP process is largely unclear and undefined.

120

Investigations of the CIP models had been conducted for the past four decades, but most of them focused on the aspect of milk fouling during dairy processing. Key efforts were made using simulated soil such as whey protein concentrates and other proteinaceous compounds to soil the milk contact surfaces; majority of which were stainless steel. The cleaning was then completed mostly using alkaline solution such as sodium hydroxide (NaOH). A multi-stage model was proposed by Harper (1972) for removal of the proteinaceous milk deposit from hard surfaces including a description of the deposit removal mechanism. Subsequently, a zeroth-order model was proposed by Schlussler (1976) for the removal of dried milk deposit. In the published literature, the first reported research using real milk deposits was for developing a mechanism- based model to describe the milk deposit removal of the holding section of a plate milk pasteurizer (Gallot-Lavallee and Lalande, 1985). The use of heated milk deposit, which simulated pasteurization process, increased the deposit mass greatly. One of the simulations based on the proposed model described a cleaning process with a deposit mass of 0.3 kg of dry matter/m2, which was orders of magnitudes higher if the depositing process was only raw milk passing through during the milking event. To simplify their models, Bird and Fryer (1991) proposed a swelling and removal two-stage model using both whey protein and whole milk as the deposits and NaOH as the cleaning agent. A range of concentrations, flow velocities, and temperatures of the cleaning solution were studied using a 6.35 mm in outer diameter stainless steel tubing. Given the more consistent data collected when using whey protein concentrates, more studies were performed using this simulated deposit to develop the foulant (i.e., deposit/soil) removal model. A three-stage cleaning process including swelling, uniform, and removal stages was proposed by

Gillham et al. (1999) using NaOH as the cleaning agent in turbulent flow conditions to remove whey protein concentrated gel deposits. The three-stage cleaning process is now widely recognized by many researchers when using NaOH to remove the proteinaceous deposits; it includes: i) the swelling stage is the penetration of the cleaning fluid into the deposit and the conversion of the deposit into a removable form; ii) the constant deposit removal rate during the 121 uniform stage, and iii) the decreasing deposit removal rate until no detectable remaining deposit on the contact surfaces during the decay stage (Xin et al., 2004).

Despite the progress made during these years, there are still knowledge gaps. Most of the studies performed used simulated soil, such as by using whey protein concentrates or other proteinaceous compounds, to simplify the experimental protocol and collect consistent experimental data; but the simulated soil does not truly represent the real raw milk deposit. In addition to proteins, other organic compounds such as lipids and inorganic compounds such as minerals in raw milk also need to be removed and these contributions need to be included in the model development. Moreover, the CIP process in these studies was not fully represented. For a complete CIP process, both alkaline and acidic solutions are needed, since they target different compounds of the deposit. Using only NaOH to conduct the CIP is not sufficient to represent the mechanism of the complete CIP process. Therefore, a more generalized mechanism-based mathematical model is needed to describe the milking system CIP. In order to achieve this, the contribution of each CIP cycle to the entire CIP process needs to be studied and understood. This can serve as the basis to optimize each cycle and/or the entire CIP process. In this study, experiments were performed using a surface evaluation simulator with stainless steel straight pipe specimens to model the deposit mass (soil) removal during the warm water rinse, alkaline wash, and acid wash cycles of the CIP process using EO water as the wash solution.

5.3 Materials and methods

5.3.1 Stainless steel surface evaluation by using the simulator

A surface evaluation simulator was specifically constructed for this study (fig. 5-1). The simulator had altogether five stainless steel straight pipe specimens each 152.4 mm in length (fig.

5-2). The stainless steel straight pipe specimen was 38.1 mm in outer and 36.8 mm in inner

122 diameters. In addition, the simulator included an initial 1.5 m glass pipe for flow and deposit visualization, followed by a 4.6 m stainless steel straight pipe as entrance length for the flow to be fully developed. To determine the needed entrance length, SolidWorks® Flow simulation module

(Version 2013, Dassault Systèmes SolidWorks Co., Waltham, MA) was used with the recommended soiling flow velocity of 0.06 m/s (ASABE, 2011). Three out of five specimens were placed as evaluation specimens for the nearer to the inlet “upstream locations” consecutively after the entrance length. The distance between the end of specimen #1 and the beginning of specimen #2, as well as the end of specimen #2 and the beginning of specimen #3, was 1.1 m, hereby denoted as recovery sections. The length of the recovery section was established based on flow perturbations induced by misalignment of rubber gaskets at connections using SolidWorks®; a 5% difference in the inner diameter was used for all the simulations (the rubber gasket shifted downward, upward, inner diameter larger or smaller cases were considered). Preliminary results showed that it took at most 0.8 m, which is much less than

1.1 m for the flow to recover for all cases. After specimen #3, there was a 1.5 m stainless steel straight pipe to straighten the flow followed by an exit glass pipe for visualization and a return elbow.

Initial studies focused only on the three specimens mentioned above, at the upstream locations, however, preliminary studies during the warm water rinse cycle showed that the designated length of the recovery sections resulted in no significant difference in the deposit soiling and removal due to the short separating distance (P>0.05). Therefore, specimens at further

(from the inlet) downstream locations were needed to find out if the distance from the inlet made a difference. Given the limited longest dimension of the available space, two additional 152.4 mm stainless steel straight pipe specimens were placed after the return elbow (fig. 5-1), designated as specimen #4 and 5 at the “downstream locations”. Specimen #4 was placed 4.6 m after the return elbow (i.e., 15.6 m from the inlet), and specimen #5 placed 1.2 m after specimen #4 (i.e., 5.8 m from the return elbow and 17.0 m from inlet). Therefore, specimens #1, 2, and 3 are hereafter 123 denoted as upstream located specimens, and specimens #4 and 5 are denoted as downstream located specimens.

Figure 5-1. Stainless steel surface evaluation simulator with specimen of 152.4 mm straight pipe

test section. [1] Inlet glass visualization, 1.5 m; [2] Entrance length, 4.6 m; [3] Upstream

locations (specimens #1, 2, and 3) of stainless steel straight pipe specimen, 152.4 mm; [4]

Recovery section, 1.1 m; [5] Return line, 1.5 m; [6] Exit glass visualization, 152.4 mm

and return elbow; [7] Downstream locations (specimens #4 and 5) of stainless steel

straight pipe specimen, 152.4 mm.

Other components essential to the milking system were also included in the simulator as previously described (Dev et al., 2014), such as a claw with liners, solution sink and receiver jar, vacuum system and milk pump. The vacuum was set at constant level of 50 kPa. A 38-L receiver

124 jar was at the end of the simulator pipeline (immediately after specimens #5) for temporary storage then either to redirect the cleaning solution back to the solution sink or to drain through the milk pump directly. The slope of the pipeline in the simulator was set to 0.8% to facilitate gravity draining.

Figure 5-2. A 152.4 mm stainless steel straight pipe specimen in the simulator [3]. Recovery

section, 1.1 m [4].

5.3.2 Experimental procedure

5.3.2.1 Number of raw milk soiling

Raw milk (~38 L) was collected from the Penn State Dairy Barn and used to soil the simulator mimicking the actual milking event occurrence. The collected milk was reheated to about 40°C (milk temperature exiting cow’s teats) in the lab to compensate the heat loss during transportation. Under vacuum, raw milk was drawn into the simulator pipeline for each run then drained without recirculation to minimize churning, cream separation, and other physio-chemical changes. Preliminary experiments were conducted using a three-time soiling procedure as

125 suggested by our previous study (Dev et al., 2014), i.e., one third (12.7 L) of the raw milk followed by a 10 min air dry, repeated three times. After the soiling was completed, the specimens were dried in an incubator (Model Symphony, VWR international, Radnor, PA), cooled in the desiccator (Model Dry-keeper AutoA-3B, Sanplatec Corp., Osaka, Japan) then weighed using a high precision balance (Model XP504, Mettler-Toledo, LLC., Columbus, OH) with a weighing range of 0 – 520 g and a readability of +/- 0.0001 g. A nondimensionalized evaluation of initial soil of deposit mass (i.e., remaining soil following a cleaning event per unit initial soil) per unit inner surface area was used for comparison since the specimens used were not identical in length (149.95±2.06 mm) considering the manufacturing tolerances. However, results showed that the deposit mass using the three-time soiling was insufficient for the potential future analysis of deposit mass removal in the alkaline wash and acid wash cycles. Given the deposit amount was largely dependent on the contact time of the raw milk with the inner (contact) surface and the specimen surface finish, five equal portions (7.6 L each time), instead of the original three equation portions (12.7 L each time), of the entire raw milk (which remained the same of 38 L) were used to maximize the soil deposit mass. It was found that the additional two soilings yielded a 10% increase in the nondimensionalized deposit mass for the initial soiling on average.

Moreover, the milk temperature during a five-time soiling process did not change much based on the time-temperature history; the milk temperature dropped from about 40°C to about 37°C after the completion of the five-time soiling, which was still acceptable.

5.3.2.2 Drying temperature and time and cooling protocol

The specimens were removed from the simulator and placed in an incubator to dry the deposits after the five-time soiling process and later on after the warm water rinse cycle, alkaline wash cycle, and acid wash cycle. The temperature of the drying process was set at 70°C, based on the following reasons: i). Aging of milk deposit is not desirable for the model development since 126 the physio-chemical properties of the milk deposits change with time; therefore a shorter drying time is favored hence a relatively higher drying temperature to speed-up the evaporation; and ii).

Previous studies had categorized the foulants of milk into two types, type A (mostly protein, formed between temperature of 75 – 110°C) and type B (mostly mineral, formed at temperature higher than 110°C) (Burton, 1968). To minimize any possibility of either foulants from forming; drying temperature of 70°C was selected.

140 130 120 110 100 90 80 Deposit amount 70 location #1 (mg) 60 location #2 50 location #3 40 30 20 10 0 0 0.5 1 2 3 5 7 Drying time (hr)

Figure5- 3. Deposit mass on the inner surfaces of upstream located specimens #1, 2 and 3 after

different drying time at 70°C in the incubator (time increment not to scale in the

horizontal axis).

A set of preliminary experiments was conducted to determine the drying time of the specimens after the soiling once the drying temperature was determined. Stainless steel straight pipe specimens were soiled as described then put in the incubator set at 70°C to dry the deposits.

127

Specimens were taken out of the incubator, cooled down to the room temperature in a desiccator then weighed after being dried for 0.5, 1, 2, 3, 5 and 7 hrs. It was determined that the weight due to the moisture loss decreased sharply in the first hour of drying (fig. 5-3), then the moisture loss slowed down, and remained constant after 5 hr of total drying time. After drying at the set time, the specimens were cooled down to room temperature in the desiccator overnight.

However, considering the same specimens were dried at 70°C in the incubator and cooled to about 25°C in the desiccator, and then used for the remaining CIP steps and dried again in the incubator at 70°C, the temperature of the specimens fluctuated frequently during the entire evaluation process, which might bring potential inaccuracies in determining the total drying time.

Therefore, upon further review, it was decided that the drying process of the specimens in this study to be 7 hr of drying time in the incubator at a drying temperature of 70°C to be sure. To balance the moisture of the deposit in the desiccator, the mass of deposit measured 24 hr after the soiling or the treatment was used as the deposited/remaining mass on the inner surface of the specimen.

5.3.2.3 Description of one complete set of experiments

In one complete set of experiments, five stainless steel straight pipe specimens were used for evaluation – specimens #1, 2, and 3 as upstream located specimens and specimens #4 and 5 as downstream located specimens. The specimens were removed after the completion of the soiling then dried in the incubator at 70°C for 7- hr. They were carefully transferred to the desiccator then left overnight for experiment next day. Meanwhile, the rest of the simulator with five dummy straight pipe sections (of same length as test specimens) inserted at the five vacated locations was shock cleaned to return the inner surfaces of the rest of the simulator to the initial state for experiment next day (Dev et al., 2014). The next day, the rest of the simulator along with the five straight pipe dummy sections, already shock cleaned, was soiled again using the same 128 five-time soiling procedure. After the rest of the simulator was freshly soiled, the five dummy straight pipe sections were removed, and the five original specimens (left in the desiccator overnight) were placed back to their respective locations. Before they were assembled into the rest of the simulator, their initial deposited mass was acquired as stated above.

The warm water rinse cycle used 40°C warm water. Due to the halfway addition of the two specimens (specimen #4 and 5) placed at the downstream locations, the warm water rinse cycle sampling time points for the upstream and downstream located specimens varied. For the upstream located specimens #1, 2, and 3, the sampling time points were 5, 20 and 30 s and for the downstream located specimens #4 and 5, the sampling time points were 10 and 30 s. Then, the alkaline wash cycle started after the completion of a 30 s warm water rinse cycle, using 70°C alkaline EO water solution. The sampling time points for the upstream located specimens #1, 2, and 3 were 30, 90, 180, and 600 s; and 30, 180, and 600 s for the downstream located specimens

#4 and 5. The test of the acid wash cycle started after the completion of a 30 s warm water rinse cycle and a 600 s alkaline wash cycle; 45°C acidic EO water solution was introduced into the simulator for 30, 90, 180, and 600 s, for specimens #1 to 5. The specific sampling time points schedule is shown in Table 5-1.

The five specimens were taken out of the simulator after treatment time sampling points and placed in the incubator for another 7 hr of drying and then cooled in the desiccator overnight.

Meanwhile, the rest of the simulator along with the five dummy straight pipe sections was shock cleaned again. After the moisture balance overnight, the residual mass on the inner surface of the specimen, namely the mass of remaining deposited soil, after the treatment sampling time point of warm water rinse cycle, alkaline wash cycle or acid wash cycle, was acquired.

129

Table 5-1. Sampling time points for the upstream and downstream located specimens during the CIP

cycles.

Sampling time points (sec)

Warm water rinse Alkaline wash Acid wash

Upstream located 0 5 20 30 30 90 180 600 30 90 180 600 specimens (#1, 2 and 3)

Downstream located 0 10 30 30 180 600 30 90 180 600 specimens (#4 and 5)

5.3.2.4 Use of nondimensionalized data

The major evaluation of this study was based on the mass of remaining deposited soil; the precision balance was accurate up to 0.1 mg of deposit on the inner surfaces of the specimens.

However, as described above, there were differences in the length of the specimens, which might cause inaccuracy in the model development. Additionally, the differences in the specimen inner surface finish influenced the amount of remaining deposit after the initial soiling and the following CIP cycles. To account for these factors, a nondimensionalized evaluation was used, taking both the difference of the specimen inner surface area and the initial soiling deposit mass into consideration. All further analysis was based on the residual deposit mass after the CIP treatment (mg) (for different sampling time points during warm water rinse cycle, alkaline wash cycle or acid wash cycle) per initial deposited after the soiling (mg) per unit inner surface area

(m2) of the specimen – the “residual mass per initial mass” takes into account the specimen inner surface finish difference, and the “mass per inner surface area” takes into account the difference in the length of the specimens.

130

5.3.2.5 Mathematical model development

Based on Reynolds number of the cleaning solution in the pipeline, the fluid flow conditions are in the turbulent regime. The simulator was designed to ensure that the turbulent flow is fully developed at the measurement locations; both at the upstream and downstream locations. Even so, the fully developed turbulent flow Navier-Stokes and energy governing equations (Kundu et al., 2012) embody several unknown parameters for our specific deposit removal study; not only during the warm water rinse but in alkaline and acidic wash cycles. Since this is the first known mechanism-based systematic analysis of deposit removal for the entire CIP process in a milking pipeline using real milk, several simplifying assumptions, consistent with and supported by the experimental results, were made. The major assumptions and justifications are as follows:

1. Beyond the fully developed entrance length, deposit removal rates during each of the cycles of the CIP process are independent of the distance from the milk inlet location including the presence of a return elbow in the pipeline.

Justification – Based on the mass deposit measurements for specimens located closer to the inlet and those located downstream, at most shared sampling times (7 out of 9), the deposit amounts were not significantly different (P<0.05). Therefore, spatial location along the milking pipeline for the simulator was not considered an independent variable. Accordingly, the mathematical formulation of deposit removal was dependent only on cycle time of the CIP process.

2. Temperature drop of the solution due to ambient conditions during each of the

CIP cycles are within an acceptable range to enable the use of mean values of solution properties for analysis.

Justification – The ambient temperature was around 25°C during the experiment period, and the average drop of cleaning solution temperature drop measured in the cleaning of milking 131 pipeline for the warm water rinse, alkaline and acid wash cycles were 0.6°C±0.15°C, 19.0°C

±0.81°C, and 5.5°C±0.92°C, respectively. The largest drop occurred during the alkaline wash cycle given the high starting temperature (70°C); the fluid density changed only 1%, from 977.8 to 988.0 kg/m3, but the dynamic viscosity increased from 0.404×10-3 to 0.547×10-3 Pa•s. The

Reynolds number correspondingly, dropped from 8×105 at the beginning of the wash cycle to

6×105 at the end of the wash cycle, but still maintained highly turbulent state compared to the turbulent flow definition with Reynolds number of greater than 5,000. Given the constant vacuum level, it is reasonable to get a more viscous fluid at the end of the wash cycle from lower temperature to be pumped at a slower, but still in the highly turbulent regime; which reflects the real-world commercial operational settings as well.

3. Deposit removal and its re-deposition at downstream location are modeled using lumped formulation.

Justification – Due to the complexity in determining the re-deposited amount, this feature was not included in the design of the first generation milking pipeline simulator.

4. Solubility of deposits is not hindered during the recycling of alkaline and acid wash solutions.

Justification – The quantity of deposits removed per liter of wash solution at the end of alkaline and acid wash cycles were at most 0.05 mg/mg/m2/L and 0.02 mg/mg/m2/L, respectively.

These are considered to be small, i.e., resulting in highly dilute solution of entrained deposits.

These assumptions simplify the mechanism-based mathematical analysis of deposit removal for each CIP cycle. Therefore, a unified mathematical model development process was used for both the upstream located (#1, 2 and 3) and downstream located specimens (#4 and 5)

(fig. 5-1).

A generalized deposit removal hypothesis applicable to each cycle was formulated; fast and slow removal of deposits occur simultaneously during each cycle, with the loosely bound 132 deposit being removed at fast deposit removal rate, but is predominant only for the beginning portion of the cycle time while the more tightly bound deposit removed at slow deposit removal rate occurs over the entire cycle time. The hypothesis will be verified for each CIP cycle based on the experimental data.

The mass deposit at any time t in the CIP is designated by m(t) and the subscript denotes the cycle of the CIP process as shown in equation 1; wherein, mTOT(t) is the total mass at time t which comprises the deposit removed during the warm water rinse (WAT), alkaline wash (ALK), and acid wash (ACD) cycles and the small amount of residual deposit (RES) after completion of the CIP process.

mTOT(t) = mWAT(t) + mALK(t) + mACD(t) + mRES(t) (1)

Where

 mTOT(t) is the total remaining deposit at any time t in the CIP process, 0 s ≤ t ≤ 1230 s,

 mWAT(t) is the deposit removed during the warm water rinse cycle, 0 s ≤ t ≤ 30 s,

 mALK(t) is the deposit removed during the alkaline wash cycle, 30 s ≤ t ≤ 630 s,

 mACD(t) is the deposit removed during the acid wash cycle, 630 s ≤ t ≤ 1230 s, and

 mRES(t) is the residual deposit amount that could not be removed even after the

completion of the EO water CIP process.

Based on the assumptions above, a unified overall first order deposit removal rate equation dependent on nth power of remaining deposit amount was proposed for all cycles as given in equation 2:

133

푑푚 (푡) 푇푂푇 = −푘 × 푚푛 (푡) (2) 푑푡 푇푂푇

The exponent n was limited to a specific integer value based on the experimental data for each CIP cycle. During the warm water rinse cycle, the deposit removal rate model was proportional to first power of mTOT(t), and correspondingly n=1; during the alkaline wash cycle and acid wash cycle however, the deposit removal rate models were proportional to zeroth power of mTOT(t), and therefore, n=0, i.e., independent of mTOT(t). Constant 푘 is the deposit removal rate specific to each CIP cycle.

5.3.2.5.1 Warm water rinse

A two-term first order deposit removal rate model was proposed for the warm water rinse cycle based upon literature review and experimental data. The first term represented the fast deposit removal rate of the loosely bound deposit and the second term represented the slow deposit removal rate contributing simultaneously of the tightly bound deposits (fig. 5-4). The rate equations were expressed as follows:

푑푚 (푡) 푊퐴푇,퐹 = −푘 × 푚 (푡) (3) 푑푡 푊퐴푇,퐹 푊퐴푇,퐹

푑푚 (푡) 푊퐴푇,푆 = −푘 × 푚 (푡) (4) 푑푡 푊퐴푇,푆 푊퐴푇,푆

Equation 3 was the deposit removal rate model of the loosely bound deposit; where,

푚푊퐴푇,퐹(푡) was the loosely bounded deposit mass at time t during the fast removal rate phase and

푘푊퐴푇,퐹 represented the fast deposit removal rate constant. Similarly, equation 4 was the deposit removal rate model of the tightly bound deposit; 푚푊퐴푇,푆(푡) was the tightly bound deposit mass under a slow deposit removal rate, represented by 푘푊퐴푇,푆.

134

Figure 5-4. Illustration of the proposed two-term deposit removal model during the warm water

rinse cycle, a fast deposit removal rate for the loosely bound deposit (0 – 30 s); however,

the contributions beyond t1 are negligible and a simultaneous slow deposit removal rate

for the tightly bound deposit (0 – 30 s).

Upon integration, the loosely and tightly bound deposit removal models could be expressed as:

(−푘푊퐴푇,퐹×푡) 푚푊퐴푇,퐹(푡) = 푚푊퐴푇,퐹(푡 = 0) × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (5)

(−푘푊퐴푇,푆×푡) 푚푊퐴푇,푆(푡) = 푚푊퐴푇,푆(푡 = 0) × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (6)

135

Where 푚푊퐴푇,퐹(푡 = 0) was defined as the maximum possible deposit removal amount for the loosely bound deposit removal (at fast deposit removal rate) and 푚푊퐴푇,푆(푡 = 0) as the maximum possible deposit removal amount for the tightly bound deposit removal (at slow deposit removal rate).The addition of 푚푊퐴푇,퐹(푡 = 0) and 푚푊퐴푇,푆(푡 = 0) was the total initial milk deposit on the specimen inner surface.

The deposit removal rate constants and maximum possible deposit removal amounts were determined from experimental data using the method of residuals (Mohsenin, 1986). In addition, the goodness of the resultant warm water rinse deposit removal rate model was evaluated based on the accuracy of the predictability of experimental data. The warm water rinse cycle in total, consisted of a simultaneous occurrence of the fast deposit removal rate for the loosely bound deposit and the slow deposit removal rate for the tightly bound deposit (fig. 5-4), which was expressed in equation 7:

푚푊퐴푇(푡) = 푚푊퐴푇,퐹(푡) + 푚푊퐴푇,푆(푡), 0 푠 ≤ 푡 ≤ 30 푠 (7)

5.3.2.5.2 Alkaline wash

During the alkaline wash cycle, based on the experimental data, the deposit removal rate model derived from equation 2 was proposed to be a zeroth power with n=0 in equations 8 and 9.

The deposit categories, however, were still separated into a distinct loosely and a tightly bound deposit as informed from experimental results.

136

Figure 5-5. Illustrations of proposed two-term zeroth power proportional model during the

alkaline wash cycle (a) and acid wash cycle (b). A fast constant deposit removal rate for

the loosely bound deposit for 30 to t2 s and a slow constant deposit removal rate from 30

s to 630 s were included in alkaline wash cycle; and a fast constant deposit removal rate

for the loosely bound deposit for 630 to t3 s and a slow constant deposit removal rate

from 630 s to 1230 s for acid wash cycle.

137

푑푚 (푡) 퐴퐿퐾,퐹 = −푘 × 푚푛 (푡), 푤𝑖푡ℎ 푛 = 0 (8) 푑푡 퐴퐿퐾,퐹 퐴퐿퐾,퐹

푑푚 (푡) 퐴퐿퐾,푆 = −푘 × 푚푛 (푡), 푤𝑖푡ℎ 푛 = 0 (9) 푑푡 퐴퐿퐾,푆 퐴퐿퐾,푆

The loosely bound deposit was removed at a fast deposit removal rate, represented by

푘퐴퐿퐾,퐹 in equation 8 and the tightly bound deposit was removed at a slow deposit removal rate, represented by 푘퐴퐿퐾,푆 in equation 9 (fig. 5-5a). The removal of the loosely bound deposit occurred from the beginning of the alkaline wash cycle to the time, t2, and then remained zero till the end of the alkaline wash cycle. However, the removal of the tightly bound deposit occurred throughout the entire alkaline wash cycle, albeit at a slow deposit removal rate. Upon integration, the loosely and tightly bound deposit removal models during the alkaline wash cycle were expressed as linear equations in equations 10 – 12:

푚퐴퐿퐾,퐹(푡) = −푘퐴퐿퐾,퐹 × 푡 + 푚푊퐴푇,퐹(푡 = 30), 30 푠 ≤ 푡 ≤ 푡2 s (10)

푚퐴퐿퐾,퐹(푡) = 0, 푡2 푠 ≤ 푡 ≤ 630 푠 (11)

푚퐴퐿퐾,푆푚퐴퐿퐾,푆(푡) = −푘퐴퐿퐾,푆 × 푡 + 푚푊퐴푇,푆(푡 = 30), 30 푠 ≤ 푡 ≤ 630 푠 (12)

The total remaining deposit after 30 s of warm water rinse was the total initial milk deposit on the specimen inner surface for the alkaline wash cycle to start with 푚푊퐴푇(푡 = 30).

138

Therefore, the mathematical deposit removal model during the alkaline wash was expressed in equation 13:

푚퐴퐿퐾(푡) = 푚퐴퐿퐾,퐹(푡) + 푚퐴퐿퐾,푆(푡) = 푚푇푂푇(푡) − 푚푊퐴푇(푡 = 30), (13) 30 푠 ≤ 푡 ≤ 630 푠

5.3.2.5.3 Acid wash

During the acid wash cycle, the deposit removal rate model derived from equation 2 was proposed to be proportional to zeroth power, i.e., n=0 in equations 14 and 15. The deposit categories were separated into the loosely and tightly bound deposits similar to those of the alkaline wash cycle.

푑푚 (푡) 퐴퐶퐷,퐹 = −푘 × 푚푛 (푡), 푤𝑖푡ℎ 푛 = 0 (14) 푑푡 퐴퐶퐷,퐹 퐴퐶퐷,퐹

푑푚 (푡) 퐴퐶퐷,푆 = −푘 × 푚푛 (푡), 푤𝑖푡ℎ 푛 = 0 (15) 푑푡 퐴퐶퐷,푆 퐴퐶퐷,푆

The loosely bound deposit was removed at a fast deposit removal rate, represented by

푘퐴퐶퐷,퐹 in equation 14 and the tightly bound deposit was removed at a slow deposit removal rate, represented by 푘퐴퐶퐷,푆 in equation 15 (fig. 5-5b). The removal of the loosely bound deposit occurred from the beginning of the acid wash cycle to the time, t3, and then remained zero to the end of the acid wash cycle. The removal of the tightly bound deposit, however, occurred along the entire acid wash cycle, at a slow deposit removal rate. Following integration, the loosely and tightly bound deposit removal models for the acid wash cycle were expressed as linear equations:

푚퐴퐶퐷,퐹(푡) = −푘퐴퐶퐷,퐹 × 푡 + 푚퐴퐿퐾,퐹(푡 = 630), 630 푠 ≤ 푡 ≤ 푡3 s (16)

푚퐴퐶퐷,퐹(푡) = 0, 푡3푠 ≤ 푡 ≤ 1230 푠 (17)

139

푚퐴퐶퐷,푆(푡) = −푘퐴퐶퐷,푆 × 푡 + 푚퐴퐿퐾,푆(푡 = 630), 630 푠 ≤ 푡 ≤ 1230 푠 (18)

Where 푚퐴퐿퐾,퐹(푡 = 630) and 푚퐴퐿퐾,푆(푡 = 630) were milk deposits on the specimen inner surface for the remaining loosely bound deposit (at fast deposit removal rate) after 30 s of warm water rinse and 600 s of alkaline wash, and the remaining tightly bound deposit (at slow deposit removal rate) after 30 s of warm water rinse and 600 s of alkaline wash, respectively. The total remaining deposit after 30 s of warm water rinse and 600 s of alkaline wash was the total initial milk deposit on the specimen inner surface for the acid wash cycle to start with, 푚퐴퐿퐾(푡 = 630).

Therefore, the net deposit removal model during the acid wash was expressed in equation 19:

mACD(t) = mACD,F(t)+mACD,S(t)=mTOT(t)–mWAT(t=30)–mALK(t=630), 630 s ≤t ≤ 1230 s (19)

5.3.2.5.4 Residual deposit expression

Previous studies have already stated the existence of a cleaning time threshold –longer cleaning time duration would result in a relatively better deposit removal, but there is always a trace of soil never removed after a threshold of cleaning time (Kulkarni et al., 1974). Therefore, it was proposed that after the completion of EO water CIP cycles, i.e., 30 s of warm water rinse,

600 s of alkaline wash and 600 s of acid wash, a residual deposit remained, which was expressed in equation 20:

mRES(t) = mACD(t)–mACD(t=1230) = mALK(t)–mALK(t=630)–mACD(t=1230) (20) = mTOT(t)–mWAT(t=30)–mALK(t=630)–mACD(t=1230)

The coefficients of the kinetic models, such as the deposit removal rates during the warm water rinse cycle (kWAT,F and kWAT,S), the slopes and intercepts of the alkaline and acid wash

140 cycles (kACD,F and mALK,F, kACD,S and mALK,S, kACD,F and mALK,F, kACD,S and mALK,S) were determined from the experimental data.

5.3.2.5.5 Validation of the developed mathematical models

To validate the developed mathematical models for the raw milk deposit removal, sampling times different from those used in model development were selected. Based on the hypotheses of the warm water rinse cycle, the alkaline wash cycle and the acid wash cycle, the raw milk deposit was removed at faster rates at the beginning of their respective cycles; therefore, the validations of the developed mathematical models were conducted during the fast deposit removal phase of the cycles. For the warm water rinse cycle, the validation time was chosen at 2 s; for the alkaline wash cycle, the validation time was chosen at 165 s and for the acid wash cycle, the validation time was chosen at 765 s.

5.3.2.6 Alternative ATP bioluminescence method

In addition to the weight measurement of the deposit, which would be a declining number working toward zero as the CIP process proceeded, an alternative approach using the ATP bioluminescence method for evaluation was used also. The inner surfaces of the specimens, after weighing for the deposited mass, were swabbed using ATP swabs and relative light unit (RLU) readings were recorded as an indication of the presence of soil. Previous studies had shown the efficacy and accuracy of using RLU to evaluate the milking system cleanliness (Dev et al., 2014;

Wang et al., 2013).

141

5.3.2.7 Visual evaluation of the specimen inner surface morphology

Specimens were also prepared for the qualitative evaluation of the inner surface morphology after the initial soiling, the warm water rinse cycle, the alkaline wash cycle, and the acid wash cycle. The specimens were soiled and washed using the same protocol as stated above.

After acquiring the initial deposited mass and the remaining deposited mass on the inner surfaces of the specimens, the specimens were immediately taken to have their inner surface morphology scanned for visual evaluation in order to eliminate excessive moisture loss in the deposit. The deposit morphology and relative amounts were observed visually using scanning electron microscopy (NanoSEM 630, Nanolab Technologies, Milpitas, CA).

5.3.2.8 Statistical analysis

Three replications were conducted for each condition. Statistical analysis was performed using Minitab® Statistical Software (Version 16.2, Minitab Inc, State College, PA). The significant differences in mean values were determined using Tukey’s method at the 95% confidence interval.

5.4 Results and discussions

A set of stainless steel straight pipes as specimens was used to study the raw milk deposit removal during the warm water rinse cycle, the alkaline wash cycle and the acid wash cycle of the milking system CIP and proposed mechanism-based deposit removal rate mathematical models.

The deposit removal rate models were developed based on the nondimensionalized deposit mass on the specimen inner surfaces, which are given in equations 7, 13, and 19 for the warm water rinse cycle, the alkaline wash cycle and the acid wash cycle, respectively.

142

5.4.1 Statistical comparisons

Initial statistical analyses of the nondimensionalized deposit mass were conducted for all the upstream located specimens (specimen #1, 2, and 3) and downstream located specimens

(specimen #4 and 5) at all CIP sampling time points. Results showed that there were no significant differences among the three upstream located specimens despite the separation of the recovery length (P>0.05). Therefore, the three upstream located specimens (specimen #1, 2, and

3) were combined and treated as “upstream locations”; similarly, no significant difference was observed between the downstream located specimens (specimen #4 and 5) (P>0.05) and therefore they were combined and treated as “downstream locations”.

During the warm water rinse cycle, it was found that for the upstream locations, there were significant differences at the initial three sampling time points, i.e., 0, 5, and 20 s of warm water rinse (P<0.05); but no significant difference was observed between the last two sampling time points, i.e., 20 and 30 s (P>0.05). Similarly, for the downstream locations, there was significant difference at the initial two sampling time points, i.e., 0 and 10 s of warm water rinse

(P<0.05) but no significant difference between the last two sampling time points, i.e. 10 and 30 s

(P>0.05).

During the alkaline wash cycle, for both the upstream and downstream locations, there were no significant differences for the last two sampling points, i.e. the 180 and 600 s sampling time points (P>0.05); during the acid wash cycle, no significant differences were observed at any sampling time points for either upstream or downstream locations despite the average decreasing trend of the deposit mass (P>0.05).

143

5.4.2 Mathematical model development

5.4.2.1 Warm water rinse cycle

The deposit removal rate model proposed for the warm water rinse cycle was a two-term exponential decaying model. Characteristic time is used to facilitate analysis; it is a measure of the time for removing 63.2% ((1-e-1)×100%) of the deposits; in the discussion of the deposit removal process, a shorter characteristic time represents an easily removed deposit while a longer characteristic time represents a tightly bound deposit which is more difficult removed. It is easily seen that at the end of the soiling process, the inner surface of the specimen was heavily soiled with milk deposit, which consists of both organic and inorganic components (fig. 5-6a).

With the introduction of the warm water, the loosely bound deposit removal occurred quickly for a short time duration (corresponding to a short characteristic time); they were removed as large bulk pieces under mainly the shear force of water under the warm temperature which facilitates to melt the fats and allow better solution/deposit contact (Alfa-Laval, 1995).

Proteins are insoluble in water, however, when adsorbed water, the protein deposits swell and became easier to be removed from the shear force of the solution for the loosely bound deposits

(Christian, 2003). The smaller and finer granule and particulate deposits, which were more tightly bound to the surfaces, were removed at a much slower rate, and the corresponding characteristic time was longer. The coefficient of the exponential term could be physically interpreted as the maximum possible deposit removal amount given that at time 0, there was no removal of deposit at all (i.e., only the initial soil existed), and the mass of that initial soil is equal to the coefficient, which is the most amount of deposit that could possibly be removed.

144

Figure 5-6.Specimen inner surface morphology at the end CIP cycles. (a) after the initial soiling

at upstream locations; (b) after the warm water rinse cycle at upstream locations; (c) after

the alkaline wash cycle at upstream locations; (d) after the acid wash cycle at upstream

locations; (e) after the shock cleaning at upstream locations; (f) after the warm water

rinse cycle at downstream locations.

For the upstream locations, from experimental data the coefficients of the loosely bound deposit removal term was calculated as 50.91 mg/mg/m2 (i.e., the maximum possible deposit removal amount for the loosely bound deposit) and the corresponding short characteristic time was 0.4 s; the coefficients of the tightly bound deposit removal term was calculated as 6.73 mg/mg/m2 (i.e., the maximum possible deposit removal amount for the tightly bound deposit) and the corresponding characteristic time was much longer, 30.0 s. The resulting mass deposit removal model is given by equation 21 and 22.

145

(−푡/0.4) 푚푊퐴푇_푈푝푠푡푟푒푎푚,퐹(푡) = 50.91 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (21)

(−푡/30.0) 푚푊퐴푇_푈푝푠푡푟푒푎푚,푆(푡) = 6.73 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (22)

The loosely bound deposit term had a characteristic time less than 1 s, indicating that the deposits, possibly large in size and loosely connected with each other, were rapidly removed once the shear force of the warm water was introduced. The removal occurred quickly at the very beginning of the warm water rinse cycle, and it was observed that within 5 s of contact with the warm water, about 90% of the initial deposit was removed; after 30 s of contact with the warm water, more than 95% of the initial deposit was removed (fig. 5-7a). This corresponded to the previous observations of the significance of the warm water rinse during the milking system CIP

(Monken and Ingalls, 2002). The difference in the maximum possible deposit removal amount between the loosely and tightly bound deposit also showed that the majority of deposit, mainly loosely bound, was removed in a large quantity at a fast deposit removal rate while the tightly bound deposit, most likely smaller in size and amount, was removed at a slow deposit removal rate. Figure 5-9a showed that the deposit removal rate decreased sharply (two orders of magnitude) within the initial 20 s, then gently decreased till the end of the warm water rinse cycle. As stated above, no significant difference (P>0.05) was observed between the last two sampling time points of 20 and 30 s warm water rinse; this indicated the 30 s warm water rinse might not be necessarily needed for a small scale milking system such as the simulator used in this study. At the end of the warm water rinse cycle, the majority of the deposit was removed and the remaining deposits left on the specimen inner surface of the upstream locations was less populated and more scattered (fig. 5-6b); similar to the structure seen in previous literature

(Gillham et al., 1999), suggesting that the remaining deposit after the warm water rinse is most likely predominantly protein and applying only warm water was not sufficient to remove the deposit on the inner surfaces.

146

Figure 5-7. Nondimensionalized experimental deposit weight change during the CIP process with

the proposed model for the upstream locations. (a) warm water rinse cycle close-up, (b)

alkaline wash cycle close-up, (c) acid wash cycle close-up, (d) CIP process including all

cycles. Hyperbolic tangent function was applied at the adjacent time of 210 s (205 – 215

s) during the alkaline wash cycle to avoid the singular point occurrence at the connection

of two straight lines. Similarly, hyperbolic function was applied at the adjacent time of

810 s (805 – 815 s) during the acid wash cycle to avoid the singular point occurrence at

the connection of two straight lines.

For the downstream locations, the coefficients of the loosely bound deposit term was calculated to be 50.41 mg/mg/m2 as the maximum possible deposit removal amount with the corresponding characteristic time was 0.9 s; the coefficients of the tightly bound deposit term was

147 calculated as 6.54 mg/mg/m2 as the maximum possible deposit removal amount with the corresponding characteristic time was 55.0 s (eqs. 23 and 24).

(−푡/0.9) 푚푊퐴푇_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = 50.41 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (23)

(−푡/55.0) 푚푊퐴푇_퐷표푤푛푠푡푟푒푎푚,푆(푡) = 6.54 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (24)

Similar to the observations from the upstream locations, the loosely bound deposit term had a characteristic time less than 1 s, suggesting that in spite of longer distance from the inlet, the deposit removal mechanism behaved similarly, i.e., the loosely bound deposit was removed rapidly as compared to the more tightly bound deposit. The maximum possible deposit removal amount of the loosely and tightly bound deposit was similar to those of the upstream locations, and this indicated an acceptable soiling process along the entire simulator pipeline.

However, from the SEM images, the observed morphology of remaining deposits was different after the warm water rinse cycle for the downstream locations (fig. 5-6f).The initial soiled deposit was indeed removed by the power of warm water and leaving a similarly less populated and more scatter deposit structure, but only to a certain extent. Different from the observation acquired from the upstream locations after the warm water rinse, similar beehive like structure was observed on the specimen inner surface at downstream locations at the completion of the warm water rinse cycle. This suggested that during the relatively low temperature warm water rinse cycle (40°C) and the no-recirculation condition, there was the existence of soil redeposition. The warm water picked up the soil at upstream surfaces nearer to the inlet, then after shifting flow direction at the return elbow and reaching the downstream surfaces further from the inlet, part of the suspended soil re-deposited on the specimen inner surface.

Statistical analysis did not show significant difference (P>0.05) between the last two sampling time points of 10 and 30 s warm water rinse in the deposit removal – indeed, the additional 20 s of warm water rinse for the downstream locations only resulted in an additional

148 about 3% of deposit removal, which could be considered minimal compared to the 90% of deposit removal occurring in the beginning 10 s of warm water rinse.

The 95% confidence interval plot of the nondimensionalized experimental deposit weight change during the CIP process is shown in figure 5-8 and the coefficient of variance of the nondimensionalized experimental deposit weight change during the CIP process is shown in table

5-2. Correspondingly, the nondimensionalized experimentally calculated deposit removal rate

(dm/dt) during the CIP process for upstream locations is shown in figure 5-10 and their coefficients of variances are listed in table 5-3.

Figure 5-8. Interval plot of nondimensionalized experimental deposit weight change during the

CIP process with 95% confidence interval for upstream locations.

149

Table 5-2. Coefficient of variance of the nondimensionalized experimental deposit weight per unit

contact area (mg/mg/m2) during the CIP process for upstream locations.

CV of nondimensionalized experimental deposit weight CIP Time (s) per unit contact area (mg/mg/m2)

0 0.87

5 24.20

20 28.41

30 27.71

60 50.97

120 53.45

210 61.33

630 64.83

660 50.31

720 25.72

810 74.35

1230 68.63

150

Figure 5-9. Experimentally calculated deposit removal rate and developed model proposed

deposit removal rate for upstream locations. (a) warm water rinse cycle close-up, (b)

alkaline wash cycle close-up, (c) acid wash cycle close-up, (d) CIP process including all

cycles.

151

Figure 5-10. Interval plot of nondimensionalized experimentally calculated deposit removal rate

during the CIP process with 95% confidence interval for upstream locations.

152

Table 5-3. Coefficient of variance of the nondimensionalized experimentally calculated deposit

removal rate (mg/mg/m2/s) during the CIP process for upstream locations.

CV of nondimensionalized experimentally calculated CIP Time (s) deposit removal rate (mg/mg/m2/s)

2.5 2.49

12.5 63.21

25 359.99

45 605.43

90 257.78

165 87.24

420 1397.84

645 5208.24

690 4342.18

765 235.69

1020 250.92

5.4.2.2 Alkaline wash cycle

The deposit removal process of the alkaline wash cycle behaved differently compared to that in the warm water rinse cycle. After the completion of the warm water rinse, most of the top layer loosely bound deposit was removed by the shear force of the warm water (Alfa-Laval,

1995); the deposit left was not easily removed by only water anymore; hence the need for alkaline solution. The already swollen protein deposits from the warm water rinse swell even more with the presence of the alkaline solution; additionally, the proteins are easily dissolved in the high hydroxide ion environment (Jeurnink and Brinkman, 1994). As stated above, a two-stage constant

153 rate deposit removal was proposed during the alkaline wash cycle (eqs. 8 and 9; fig. 5-5) – a fast constant deposit removal rate at the beginning of the alkaline wash cycle and a continuous slow constant deposit removal rate throughout the entire alkaline wash cycle. The slopes and intercepts of both deposit removal models were calculated from experimental data.

For the upstream locations, the fast deposit removal models were expressed as:

푚퐴퐿퐾_푈푝푠푡푟푒푎푚,퐹(푡) = −0.010566 × 푡 + 2.218726, 30 푠 ≤ 푡 ≤ 210 푠 (25)

푚퐴퐿퐾_푈푝푠푡푟푒푎푚,퐹(푡) = 0, 210 푠 ≤ 푡 ≤ 630 푠 (26)

and equation 27 showed the model for the slow deposit removal process for the upstream locations during the alkaline wash:

푚퐴퐿퐾_푈푝푠푡푟푒푎푚,푆(푡) = −0.000111 × 푡 + 0.581058, 30 푠 ≤ 푡 ≤ 630 푠 (27)

2 The fast deposit removal rate (푘퐴퐿퐾_푈푝푠푡푟푒푎푚,퐹 = -0.010566 mg/mg/m /s) shown in equation 25, corresponding to a loosely bound deposit removal process, was valid only for a short time duration at the beginning of the alkaline wash (30 to 210 s), then its contribution to the deposit removal was minimal during the wash cycle (eq. 26). The tightly bound deposit contributed throughout the entire alkaline wash cycle from the beginning to the end, albeit a

-4 really slow deposit removal rate on the order of magnitude of 10 (푘퐴퐿퐾_푈푝푠푡푟푒푎푚,푆 = -0.000111 mg/mg/m2/s) (fig. 5-9b). It is easily seen that this slow deposit removal process did not contribute much to the entire deposit removal during the alkaline wash cycle, especially after the completion of the loosely bound deposit removal at 210 s (fig. 5-7b); therefore, it is possible to project that for a small scale milking system such as the simulator used in this study, the additional time (210 to 630 s) for the slow deposit removal process might not be necessary.

At the end of the alkaline wash cycle, it was observed that the specimen inner surface had even less residual deposit, and some clear background of the specimen inner surface finish was

154 visible (fig. 5-6c). It had been reported that with the reinforcement of the alkaline solution on the protein deposit, potential cracks would form leading to an increased penetration of the wash solution into the deposit layer (Christian, 2003). The deposit observed in this study was crisper and more scattered as compared to the morphology after the warm water rinse cycle, and the smaller structure indicated that the majority of the organic deposit, typically large in size, such as protein and lipid were removed during the alkaline wash cycle as previously observed (Gillham,

1997), leaving only the crystalized inorganic deposits such as minerals on the specimen inner surface.

Similarly, for the downstream locations during the alkaline wash cycle, the fast deposit removal models were expressed in equations 28 and 29, and the slow deposit removal model was presented in equation 30. A similar trend was observed – the contribution of the loosely bound deposit removal lasted only for 180 s from the beginning of the alkaline wash cycle (eq. 28), and then its contribution to the entire wash cycle deposit removal was minimal (eq. 29). The fast

2 deposit removal rate for the downstream locations (푘퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,퐹 = -0.013488 mg/mg/m /s) was on the same order of magnitude as that of the upstream locations; also indicating a similar deposit removal mechanism between the upstream and downstream locations, which justified the assumptions proposed in the materials and methods section. For the tightly bound deposit contributing to the entire alkaline wash cycle, the deposit removal rate was on the order of magnitude of 10-7 mathematically (eq. 30); but from practical considerations this small magnitude of deposit removal could be treated as negligible.

푚퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = −0.013488 × 푡 + 2.832433, 30 푠 ≤ 푡 ≤ 210 푠 (28)

푚퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = 0, 210 푠 ≤ 푡 ≤ 630 푠 (29)

푚퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,푆(푡) = −0.0000003 × 푡 + 1.360970, 30 푠 ≤ 푡 ≤ 630 푠 (30)

155

5.4.2.3 Acid wash cycle

The acid wash cycle had a similar trend of deposit removal with that of the alkaline wash cycle, which is a fast deposit removal rate at the beginning of the wash cycle and a slow deposit removal rate throughout the entire wash cycle. However, it was in general a much slower process compared to the alkaline wash given that, as the CIP process moved on, less and less deposit was left on the specimen inner surfaces to be removed. This bottom layer had been reported to cover the surface, tightly bound, and hard to be removed (Burton, 1968). Based on previous discussions, the exact composition of this tightly bound layer including majorly protein and mineral, differed based on the constituents of the milk itself and the process treatment (Burton,

1968); such as heated surface at 85°C would result in high ash content as previously reported

(Lyster, 1965). The proteins existed in this phase is presented to be slightly acid soluble but for the minerals, most of them are acid soluble and therefore could be removed relatively more effectively using the acid wash solution (Grasshoff, 1997).

The deposit removal model for the loosely bound deposit was expressed in equations 31 and 32; the loosely bound deposit was removed at a fast rate within short time duration at the beginning of the wash cycle then remained minimal (fig. 5-7c). The fast deposit removal rate

2 (푘퐴퐶퐷_푈푝푠푡푟푒푎푚,퐹 = -0.001131 mg/mg/m /s) was almost one order of magnitude smaller than that

2 for the alkaline wash (푘퐴퐿퐾_푈푝푠푡푟푒푎푚,퐹 = -0.010566 mg/mg/m /s), indicating a harder removal process. The deposit removal rate of the tightly bound deposit (푘퐴퐶퐷_푈푝푠푡푟푒푎푚,푆 = -0.000123

2 mg/mg/m /s), however, was comparable with the corresponding alkaline wash (푘퐴퐿퐾_푈푝푠푡푟푒푎푚,푆 =

-0.000111 mg/mg/m2/s). The tightly bound deposit removal progressed at slow deposit removal rate (fig. 5-11c), but steady; this observation in turn proved that the deposit removal was getting more and more difficult; the energy/cost required became higher to get these tightly bound deposit removed. This observation was proved again when the optimization of the simulator CIP

156 was conducted based on the deposit weight measurement – indicating that a longer time of the acid wash cycle was needed to achieve a satisfactory CIP performance.

At the end of the acid wash cycle, it was clearly seen that most of the deposit was removed, and the specimen inner surface was easily seen during the SEM inspection (fig. 5-6d).

The original crisp and scattered minerals seen after the completion of alkaline wash cycle were removed during the acid wash cycle and thus confirmed the completion of the CIP cycle.

푚퐴퐶퐷_푈푝푠푡푟푒푎푚,퐹(푡) = −0.001131 × 푡 + 0.916242, 630 푠 ≤ 푡 ≤ 810 푠 (31)

푚퐴퐶퐷_푈푝푠푡푟푒푎푚,퐹(푡) = 0, 810 푠 ≤ 푡 ≤ 1230 푠 (32)

푚퐴퐶퐷_푈푝푠푡푟푒푎푚,푆(푡) = −0.000123 × 푡 + 0.384996, 630 푠 ≤ 푡 ≤ 1230 푠 (33)

For the downstream locations during the acid wash cycle, the deposit removal rate for

2 loosely bound deposit (푘퐴퐶퐷_퐷표푤푛푠푡푟푒푎푚,퐹 = -0.004885 mg/mg/m /s) was around half of that

2 during the alkaline wash cycle (푘퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,퐹 = -0.013488 mg/mg/m /s) given the harder deposit removal process. But the rate for the tightly bound deposit removal (푘퐴퐶퐷_퐷표푤푛푠푡푟푒푎푚,푆 =

-0.000226 mg/mg/m2/s) was higher compared to that of the alkaline wash cycle

2 (푘퐴퐿퐾_퐷표푤푛푠푡푟푒푎푚,퐹 = -0.0000003 mg/mg/m /s). However, both rates were on the order of magnitudes of 10-4 or lower; therefore, they could be practically considered negligible.

푚퐴퐶퐷_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = −0.004885 × 푡 + 3.956921, 630 푠 ≤ 푡 ≤ 810 푠 (34)

푚퐴퐶퐷_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = 0, 810 푠 ≤ 푡 ≤ 1230 푠 (35)

푚퐴퐶퐷_퐷표푤푛푠푡푟푒푎푚,푆(푡) = −0.000226 × 푡 + 0.623696, 630 푠 ≤ 푡 ≤ 1230 푠 (36)

5.4.2.4 Overall model summary

The overall deposit removal process, as stated above, started with a sharp deposit mass decrease at the beginning of the warm water rinse cycle. At the end of the 30 s warm water rinse,

157 more than 90% of the initial soil was removed – this suggested the importance of the timely and appropriate warm water rinse cycle for the milking system CIP. The organic deposits, removed mostly during the alkaline wash cycle, contributed to an additional 4% (the upstream locations of

3.4% and the downstream locations of 4.3%). The mineral deposits, removed mostly during the acid wash cycle, contributed to an average of additional 1% (the upstream locations of 0.5% and the downstream locations of 1.8%); leaving about 0.5% residuals on average not removed by the entire EO water CIP process for both the upstream and downstream locations (fig. 5-7d).

The deposit removal rate decreased orders of magnitude during the warm water rinse cycle, and then stayed at different constants during the alkaline wash cycle and acid wash cycle.

The loosely bound deposits were removed at faster rates compared to the tightly bound deposits; at both the upstream and downstream locations and for both alkaline wash cycle and acid wash cycle. As CIP process progressed, the deposit removal turned harder and harder, leading to a slower removal rate and potential longer CIP time duration (fig. 5-11d).

To evaluate the goodness of fit of the developed models, root mean square error (RMSE) and percent error difference were used. RMSE is used to estimate the accuracy of the developed model by comparing the differences between the predicted values from the developed model with the experimentally acquired data. The percent error difference, inspired from the definition of coefficient of variation, evaluates the differences between the experimentally acquired data and the RMSE (eq. 37).

Percent Error Difference = (RMSE / Experimental deposit mean value)*100% (37)

The percent error differences at the upstream and downstream locations for the individual

CIP cycles and overall CIP were summarized in table 5-4. For the upstream locations, the developed deposit removal model resulted in a 0.23 mg/mg/m2 RMSE and an overall 3.67% of the percent error difference (fig. 5-7d). The alkaline wash cycle and acid wash cycle had higher percent error differences, 4.20% and 3.97%, respectively, as compared to the warm water rinse 158 cycle. This indicated that as the CIP process moved on, the decreasing deposit weight measurement resulted in higher fluctuations compared to the warm water rinse cycle. For the downstream locations, the developed deposit removal model resulted in a 0.07 mg/mg/m2 RMSE and an overall 0.93% of the percent error difference. Similarly, as compared to the warm water rinse cycle, the alkaline and acid wash cycles had higher percent error differences. However, despite the overall lower percent error difference for the downstream locations, the acid wash cycle itself, was really high in the evaluation. Possible explanations might come from the extremely small deposit amount left to be measured during the acid wash cycle and the relative higher fluctuations in the experimental measurement. The distinct behavior in the percent error difference between the upstream and downstream locations, especially during the alkaline wash cycle and acid wash cycle, indicated that combining the models of the upstream and downstream locations does not seem to be reasonable.

Table 5-4. Percent error difference during different CIP cycles for the upstream and downstream

locations

Percent error difference (%)

Warm water Alkaline wash Acid wash Overall CIP rinse cycle cycle cycle

Upstream locations 1.90 4.20 3.97 3.67

Downstream locations 0.25 0.60 11.16 0.93

The validations of the developed models were conducted at 2 s for the warm water rinse cycle, 165 s for the alkaline wash cycle and 765 s for the acid wash cycle. The term percent variation was used for comparison. It considers the differences between the values from the

159 model prediction with the experimental acquired data as compared to the experimental data, and expressed in equation 38.

(푀표푑푒푙 푝푟푒푑푖푐푡푖표푛 −퐸푥푝푒푟푖푚푒푛푡푎푙 푑푎푡푎) Percent Variation = × 100% (38) 퐸푥푝푒푟푖푚푒푛푡푎푙 푑푎푡푎

The percent variations at the upstream and downstream locations for the CIP cycles were summarized in table 5-5. The average percent variation for the upstream and downstream locations was 0.84% for the warm water rinse cycle, 3.41% for the alkaline wash cycle and -

2.29% for the acid wash cycle. It is seen that the model predicted values match with the experimental data well with low percent variation, indicating that the developed models were successfully validated at these time points. The percent variation not only reveals how close the developed models are capable to predict the experiments, but also informs what trend these models predict which could shed light on the further improvements and adjustments of the models. Sometimes, the developed model might be over-prediction or under-prediction all the time, and if that happens, it is useful information for the developers to adjust the models accordingly. However, as is shown in the table 5-5, no clear trend of over- or under-prediction was observed in the developed models in this study. The developed models for the upstream locations over-predicted slightly during the warm water rinse cycle, but under-predicted during the alkaline wash and acid wash cycles. However, for the downstream locations, the developed models under-predicted slightly during the warm water rinse cycle, but over-predicted during the alkaline wash and acid wash cycles. Therefore, it is concluded that the developed models were validated at the selected time points and no further improvements and adjustments were needed to correct the potential biased developed models.

160

Table 5-5. Percent variation during different CIP cycles for the upstream and downstream locations

Percent variation (%)

Warm water rinse cycle Alkaline wash cycle Acid wash cycle

Upstream locations 2.55% -1.80% -4.73%

Downstream locations -0.88% 8.61% 0.15%

Despite the minute differences discussed above, the overall evaluation of the developed models under these comparisons suggested that the errors from the experimental data with the developed models did not deviate much and therefore the developed models could be considered acceptable.

5.4.3 ATP bioluminescence validation

The results from the weight-based mathematical model were also validated using the ATP bioluminescence method. Results from the ATP bioluminescence method were original expressed in Relative Light Unit (RLU) readings, but the log10 of RLU+1 was used for comparison purpose and to avoid the indeterminate value for log10 of zero; i.e., since RLU of zero is a possible outcome. Higher RLU represents a dirtier surface and lower RLU represents a cleaner surface; ideally RLU of 0 is desirable to represent a thoroughly cleaned surface. The sensitivity of ATP bioluminescence had been demonstrated by several workers (Cais and Pikul, 2008; Samkutty et al., 2001; Vilar et al., 2008) and hereby used as an alternative to test the cleanliness of the EO water CIP for the simulator.

161

5.4.3.1 Original CIP process validation

The original CIP process, described as a 30 s warm water rinse, followed by a 10 min alkaline wash and a 10 min acid wash, was first tested using the ATP bioluminescence method at the completion of each CIP cycle. The results, shown in figure 5-11, demonstrated that the initial milk deposit, even non-inoculated with microorganisms (Wang et al., 2014), was abundant for ready detection by ATP; i.e., order of magnitude of 106. After a 30 s warm water rinse, the RLU reading reduced to an average of 105 for both the upstream and downstream locations, and the

RLU reading further reduced to 101 for the upstream locations and 102 for the downstream locations at the end of the alkaline wash cycle. At the end of the acid wash cycle, the RLU was not the ideal reading of 0, but based on the cutoff RLU reading for the stainless steel material, the surface could be defined as “clean” for both the upstream and downstream locations (Wang et al.,

2013). It was observed here again, that there were differences between the RLU readings for the upstream and downstream locations, as found in the above nondimensionalized deposit mass measurements. Nonetheless, despite the higher RLU readings when compared to the upstream location readings, the downstream location RLU results were still below the cutoff RLU reading, which indicated that even for these specimens, located relatively further from the inlet, was still considered as “clean” based on the manufacture’s recommendations.

When examining the RLU reduction percentage, it was found that, at the end of the warm water rinse cycle, about 97.76% of ATP was removed compared to the initial ATP, which validated the results from the weight measurement (an average of 94.55%). At the end of the alkaline wash cycle, 99.999% of the initial ATP was removed (an average of 98.4% from the weight measurement); and at the end of the acid wash cycle, 99.9997% of the initial ATP was removed (an average of 99.55% from the weight measurement).

162

( )

Figure 5-11. Nondimensionalized deposit mass and log(RLU+1) at the end of the original CIP

cycles of 30 s warm water rinse, 10 min alkaline and 10 min acid wash. (a) upstream

locations, (b) downstream locations.

163

5.4.3.2 CIP process validation

From the developed weight-based mathematical model, a series of CIP experiments were conducted by shortening the CIP cycle times. From the deposit removal process of different CIP cycles, it could be concluded that for both the upstream and downstream locations, after 10 s of warm water rinse, more than 90% of the initial deposited soil (the majority of which was loosely bound) was removed by the shear force of the warm water; during the alkaline wash and acid wash cycles, the deposit removal rate decreased after the initial 180 s of the wash cycle.

Therefore, it was proposed to shorten the CIP cycles from the original 30 s warm water rinse to

10 s of warm water rinse, from the original 10 min alkaline wash to 3 min (180 s) of alkaline wash and from the original 10 min acid wash to 3 min (180 s) of acid wash, respectively.

Therefore, if effective in cleaning, the overall CIP process time would be reduced by 70% - from

20.5 to 6.17 min. From the weight measurement and RLU results (fig. 5-12), it was observed that under the newly revised protocol, at the end of the 3 min acid wash cycle, the remaining deposit on the inner surfaces of the specimens was higher than the original 30 s/10 min/10 min CIP; and the same trend was also observed from the high RLU readings, suggesting a need for improvement. It was observed that, at the end of the 3 min alkaline wash, the remaining deposit for both the upstream and downstream locations was comparable to the original 30 s/10 min/10 min CIP - 0.51±0.31 mg/mg/m2 for the original CIP vs. 0.71±0.20 mg/mg/m2 for the proposed

CIP for the upstream locations and 1.36±0.75 mg/mg/m2 for the original CIP vs. 1.64±0.51 mg/mg/m2 for the proposed CIP for the downstream locations. However, no similarly comparable results were found at the end of the acid cycle for the upstream and downstream locations -

0.23±0.15 mg/mg/m2 for the original CIP vs. 0.56±0.16 mg/mg/m2 for the proposed CIP for the upstream locations and 0.35±0.13 mg/mg/m2 for the original CIP vs. 0.91±0.30 mg/mg/m2 for the proposed CIP for the downstream locations. This indicated that it was the 3 min acid wash that was not sufficient to achieve a satisfactory CIP performance. This was reasonable given the

164 discussions above of the slower deposit removal process for the acid wash cycle, as compared to that of the alkaline wash cycle. Therefore, an extended time for the acid wash cycle was used as the second optimization trial– the acid wash time was extended from 3 min to 6 min (fig. 5-13).

Results showed that the additional 3 min acid wash achieved a better cleaning performance, and the remaining deposit mass was comparable to the original 30 s/10 min/10 min CIP - 0.23±0.15 mg/mg/m2 for the original CIP vs. 0.24±0.18 mg/mg/m2 for the revised CIP for the upstream locations and 0.35±0.13 mg/mg/m2 for the original CIP vs. 0.32±0.18 mg/mg/m2 for the revised

CIP for the downstream locations. Even though slightly longer than first proposed, this revised

CIP process still resulted in shortening the original CIP process time by 55% from 20.5 to 9.17 min and therefore saving time and energy consumption.

It was observed that the RLU reading at the end of the 3 min or 6 min acid wash cycle did not alter much (fig. 5-13). Possible reason might come from the limited number of samples taken and the potential aging of the stainless steel surface harboring residual therefore affecting the

RLU reading. However, what was confirmed experimentally was that at the end of the 10 s/3 min/6 min CIP, the RLU readings for both the upstream and downstream locations were below the cutoff RLU reading, and therefore, the revised CIP process using EO water did accomplish acceptable level of cleaning.

165

( )

Figure 5-12. Nondimensionalized deposit mass and log(RLU+1) at the end of the optimized CIP

cycles of 10 s warm water rinse, 3 min alkaline and acid wash. (a) upstream locations, (b)

downstream locations.

166

( )

Figure 5-13. Nondimensionalized deposit mass and log(RLU+1) at the end of the further

optimized CIP cycles of 10 s warm water rinse, 3 min alkaline and 6 min acid wash. (a)

upstream locations, (b) downstream locations.

167

Another observation was the relationship between the RLU reading and the remaining nondimensionalized deposit mass. For both the upstream and downstream locations, there was a near-linear relationship between the log (RLU+1) and the remaining nondimensionalized deposit mass when taking out the initial nondimensionalized deposit mass and the initial RLU readings

(fig. 5-14). For both locations, the coefficient of determination, R2, was greater than 85%. This observation was of significance because it provided insight that using ATP bioluminescence method could be a rapid, yet precise, alternative method to understand how much the residual weight was on the specimen after a certain CIP cycle without the need for an accurate balance. At lower levels of measurements for both the upstream and downstream locations, greater deviations from the fit lines existed. Possible explanations would be the precision of the ATP measurements

– with the capacity of detecting as high as 108 of RLU readings, at lower levels of presence of soil, the detection might be less sensitive.

Minor differences existed in the relationships of the upstream and downstream locations, which were reflected from the slopes and intercepts of the regression equations and the coefficient of determination. This suggested that lumping the model of the upstream locations with the mode of the downstream locations is not recommended. Upstream and downstream locations have distinct characteristics due to the different distances from inlet and the presence of a return elbow in the simulator setup. At the end of the shock cleaning, the specimen inner surface morphology was different as compared to the completion of the EO water CIP process (fig. 5-6e).

Possible reasons might come from the application of the double concentrated chemicals during the shock cleaning process.

168

Figure 5-14. Linear relationship between the nondimensionalized deposit mass and the

log(RLU+1) at the end of the further optimized CIP cycles of 10 s warm water rinse, 3

min alkaline and 6 min acid wash, without the initial nondimensionalized deposited mass

and initial ATP RLU reading. (a) upstream locations, (b) downstream locations.

169

5.5 Conclusions

This study was undertaken to quantify the milk deposit removal rate of the milking system CIP using EO water through a surface evaluation simulator of stainless steel straight pipes. Stainless steel straight pipe specimens were evaluated by weighing the mass deposited on the inner surfaces after a five-time soiling process and the amount remaining at several sampling time values during each of the warm water rinse, alkaline wash, and acid wash cycles of the CIP process. Additional validation was conducted using the ATP bioluminescence method. Based on deposit removal analysis, the CIP process duration was shortened. Results showed that the warm water rinse cycle contributed the most to the CIP process; a 10 s warm water rinse resulted in more than 90% of the initial deposited soil removed, for both the upstream and downstream locations. A two-term exponential model was proposed for the warm water rinse cycle to describe the loosely bound bulk deposit and the tightly bound granule deposit removal process. The deposit removal during the alkaline and acid wash cycles progressed at constant rates and behaved similarly - the loosely bound deposit was removed at a fast deposit removal rate lasting for only the beginning 180 s of the wash cycle and the tightly bound deposit was removed at a slow deposit removal rate throughout the entire wash cycle. Detailed specimen inner surface morphology was observed using scanning electron microscopy to confirm the progress of each

CIP cycle. The CIP for the simulator was shortened by 55% (from 20.5 to 9.17 minutes) based on the proposed models from the original 30 s warm water rinse/10 min alkaline wash/10 min acid wash to 10 s warm water/3 min alkaline wash/6 min acid wash to achieve a satisfactory CIP performance. Additional studies were conducted using ATP bioluminescence method for the deposit mass measurement validation and the relationship between the RLU readings and the deposit mass exhibited a near-linear regression. In conclusion, this is the first study to evaluate the deposit removal rate during the EO water CIP process as an alternative CIP process.

170

5.6 Acknowledgements

Funding for this project was provided in-part by a USDA Special Research Grant (No.

2010-34163-21179) and the Pennsylvania Agricultural Experiment Station. We also would like to acknowledge Dr. Roderick Thomas, Randall Bock, and all Penn State dairy barn personnel for their help in the project.

5.7 References

Alfa-Laval, A.B. (1995). Dairy Handbook. Alfa-Laval, Lund, Sweden.

ASABE Standards. (2011). AD5707:2007: Milking machine installations – Construction and

performance. St. Joseph, Mich.: ASABE.

Bird, M.R., & Fryer, P.J. (1991). An experimental study of the cleaning of surfaces fouled by

whey proteins. Food Bioprod Proc., 69,13-21.

Burton, H. (1968). Deposits from whole milk in heat treatment plant-a review and discussion. J

Dairy Res., 35,317–330.

Cais, S. D., & Pikul, J. (2008). Evaluation of steel surface cleanliness level in dairies using the

bioluminescence method. Bull Vet Inst Pulawy., 52, 625–629.

Christian, G. K. (2003). Cleaning of carbohydrate and dairy protein deposits. PhD diss.

Birmingham, UK: University of Birmingham, Department of Chemical Engineering.

Dev, S.R.S., Demirci, A., Graves, R.E., & Puri, V.M. (2014). Optimization and modeling of an

electrolyzed oxidizing water based Clean-In-Place technique for farm milking systems

using a pilot-scale milking system. J. Food Eng. 135,1–10.

DPC. (2010). Number 4: Guidelines for installation, cleaning, and sanitizing of large and multiple

receiver parlor milking systems. Richboro, PA.:Dairy Practices Council.

171

Gallot-Lavallee, T & Lalande, M. (1985). A mechanistic approach of pasteurized milk deposit

cleaning. In Proc. of the 2nd Int. Conf. on Fouling and Cleaning in Food Processing, (pp.

374-394). Madison, WI.

Gillham, C.R., Fryer, P.J., Hasting, A.P.M., & Wilson, D.I. (1999). Cleaning-in-place of whey

protein fouling deposits: mechanisms controlling cleaning. Food Bioprod Proc.,77,127-

136.

Grasshoff, A. (1997). Cleaning of heat treatment equipment. International Dairy

Federation (IDF) monograph, Fouling and cleaning in Heat Exchangers.

Harper, W. J. (1972). Sanitization in dairy food plants. In Food Sanitization, (pp. 130-160).

Westport, CT: The Avi Publ. Co., Inc.

Hsu, S. Y. (2005). Effects of flow rate, temperature and salt concentration on chemical and

physical properties of electrolyzed oxidizing water. J. Food Eng., 66,171 – 176.

Jeurnink, T.J.M. & Brinkman, D.W. (1994). The cleaning of heat exchangers and evaporators

after processing milk or whey. Int. Dairy J., 4, 347 – 368.

Kulkarni, S. M., Maxcy, R. B., & Arnold, R. G. (1974). Evaluation of soil deposition and removal

processes: an interpretive review. J. Dairy Sci. 58(12), 1922 – 1936.

Kundu, P. J., Cohen, I. M., & Dowling, D. R. (2012). Fluid Mechanics. 5th edition. Elsevier,

Waltham, MA, USA.

Lyster, R. L. J. (1965). The composition of milk deposits in an ultra-high-temperature plant. J.

Dairy Res. 32. 203 – 210.

Mohsenin, N. N. (1986). Physical Properties of Plant and Animal Materials: Structure, Physical

Characteristics, and Mechanical Properties. New York.:Gordon and Breach, Science

Publishers.

172

Monken A. & W. Ingalls. (2002). Milking system cleaning and sanitizing: troubleshooting milk

bacteria counts. In National Mastitis Council Regional Meeting Proceedings, (pp. 55-60).

Kansas City, MI.

Reinemann, D. (1995). System design and performance testing for cleaning milking systems. In

NRAES Conf. Designing a Modern Milking Center: Parlors, Milking Systems,

Management and Economics. Indianapolis, Ind.: NRAES.

Samkutty, P. J., Gough, R. H., Adkinson, R. W., & McGrew, P. (2001). Rapid assessment of the

bacteriological quality of raw milk using ATP bioluminescence. J. Food Prot.., 64, 208–

212.

Schlussler, H. J. (1976). Kinetics of the cleaning action on hard surfaces. Brauwissenschaft.,

29,263-276.

Vilar, M. J., Rodríguez-Otero, J. L., Diéguez, F. J., Sanjuán, M. L., & Yus, E. (2008). Application

of ATP bioluminescence for evaluation of surface cleanliness of milking equipment.

Intl.J. Food Micro., 125, 357–361.

Wang, X., Dev, S.R.S., Demirci, A., Graves, R.E., & Puri, V.M. (2013). Electrolyzed oxidizing

water for cleaning-in-place of on-farm milking systems performance evaluation and

assessment. Appl. Eng. Agric., 29,717–726.

Wang, X., Demirci, A., Puri, V.M., & Graves, R.E. (2014). Evaluation of blended electrolyzed

oxidizing water-based Cleaning-In-Place technique using a lab scale pilot milking

system. Transactions of ASABE. (submitted)

Xin, H., Chen, X., & Ozkan, N. (2004) A mathematical model of the removal of milk protein

deposit. In ECI Conference on Heat Exchanger Fouling and Cleaning: Fundamentals

and Applications, (pp. 158 – 166). Santa Fe, NM.

173

CHAPTER 6

ONE-STEP CLEANING-IN-PLACE FOR MILKING SYSTEMS AND

MATHEMATICAL MODELING OF DEPOSIT REMOVAL FROM

STAINLESS STEEL PIPELINE USING BLENDED ELECTROLYZED

OXIDIZING WATER

6.1 Abstract

Cleaning-in-place (CIP) is widely used on dairy farms to clean and sanitize the inner surfaces of the milking system components after the milking event is completed. The conventional milking system CIP includes: i) warm water rinse; ii) alkaline wash; iii) acid wash; and iv) sanitizing rinse before the next milking event. Recently an increasing number of dairy farms are adopting a one-step CIP process, which combines the alkaline wash and acid wash cycles together as one wash cycle. This one-step CIP has the advantage of saving time and reducing energy consumption and chemical usage. Electrolyzed oxidizing (EO) water is an emerging technology producing alkaline and acidic EO water by electrolyzing dilute sodium chloride solution. Previous studies in our lab had shown that the alkaline and acidic EO water solutions, as separate solutions for a two-step CIP process, could be used as an alternative for the conventional milking system CIP. Blending the alkaline EO water with the acidic EO water at certain ratio could be used as one-step CIP process. Therefore, this study was undertaken to evaluate the deposit removal process during the one-step CIP wash cycle using the already optimized blended EO water solution and the stainless steel surface evaluation simulator.

Stainless steel straight pipes were used as testing specimens and the remaining milk deposit mass on the inner surfaces of the specimens was evaluated. A two-term exponential decay kinetic model was developed for the blended EO water one-step EO water wash with an initial fast

174 deposit removal rate along with a slow deposit removal rate throughout the entire one-step wash.

The proposed models matched the experimental data with acceptable root mean square errors (an average of 0.08 mg/mg/m2) and low percentage error differences (an average of 5.16%). The relative light unit (RLU) reading from the ATP bioluminescence method was also used as an indirect evaluation at each time sampling point during the one-step wash cycle. Furthermore, scanning electron microscopy (SEM) was used to understand the residual deposit morphology on and qualitatively evaluate cleanliness of the inner surfaces of the specimens after the one-step wash. Further deposit coverage reduction, on average, of 60% of the viewing area after the blended EO water one-step wash was observed as compared to the morphology after the warm water rinse cycle. Moreover, results showed that the specimen at the completion of the blended

EO water one-step wash could be considered as clean as indicated from the RLU cutoff reading of stainless steel material and the developed mathematical model for the blended EO water one- step wash was acceptable for the simulator application.

6.2 Introduction

On a commercial dairy farm, the cleaning and sanitizing of the milking system including the transportation and processing pipelines and storage facilities was typically done using a highly automated cleaning-in-place (CIP) process. The conventional milking system CIP consists of four cycles: i) warm water rinse; ii) alkaline wash; iii) acid wash; and iv) sanitizing rinse before the next milking event. Electrolyzed oxidizing (EO) water is a technology which generates alkaline and acidic EO waters through the electrodialysis of a dilute salt solution. Under certain conditions, the properties of the EO water solution fit well with the requirements of the conventional milking system CIP. The CIP performance of EO water has been compared to the conventional CIP and results were satisfactory with a reduced operational cost in our earlier studies (Dev et al., 2014).

175

Recently, a new CIP approach referred as “one-step” CIP has received attention for dairy farms, which combined the conventionally separated alkaline and acid wash cycles into one (Parr,

2013). There are commercially available one-step CIP chemicals for purchase and the CIP performance was claimed to be comparable to the conventional CIP. The advantage of the one- step CIP was a saving in water and chemical usage including reduction in energy consumption and wash time. Since alkaline and acidic fluid streams emerge from the EO system, by redirecting partial alkaline EO water into the acidic EO water, a blended EO water solution could be produced. With the relatively less corrosive pH and high ORP, blended EO water solution was evaluated by several studies (Guentzel et al., 2008; Guentzel et al., 2010; Guentzel et al., 2011).

The disinfecting effectiveness of the blended EO water at near neutral pH had been demonstrated on produce such as vegetables and fruits (Guentzel et al., 2010), and on food processing plants

(Guentzel et al., 2011). Based on these studies, it was proposed that the blended EO water at certain blending ratio could be used as an alternative for the one-step milking system CIP. In

Chapter 4, the one-step CIP performance using the blended EO water solution was conducted on a lab scale pilot milking system and optimized the parameters affecting the CIP process. Results showed that cleaning time duration of 17 min, a starting temperature of 59°C, and an acidic EO water percentage of 60% in the blended EO water solution could achieve an optimal CIP performance on the sampling locations of stainless steel elbows and pipes. Moreover, using the blended EO water at the optimal condition achieved comparable CIP performance when compared to the commercially available one-step CIP chemicals at a much lower operational cost.

There had been studies investigating the CIP through modelling of each CIP cycle and demonstrate the milk soil deposition and removal mechanism. Most of the studies, however, used simulated soil such as whey protein concentrate as the deposited soil and sodium hydroxide as the

CIP solution during the model development process to acquire consistent experimental data.

Several models were proposed by researchers (Harper, 1972; Schlussler, 1976; Gallot-Lavallee

176 and Lalande, 1985; Bird and Fryer, 1991; Gillham et al, 1999) and a widely recognized model using NaOH to remove the proteinaceous deposits consists of three stages: i) the swelling stage when the cleaning fluid penetrates into the deposit and convert the deposit matrix into a removable form; ii) the uniform stage when the deposit removal rate remains constant, and iii) the decay stage when the deposit removal rate starts to decease until there is no detectable remaining deposit on the deposit contact surfaces (Xin et al., 2004).

The knowledge gap of these studies was the lack of reality when applied to the milking system CIP. Using simulated soil does not fully represent the residual milk constituents and using only NaOH as the cleaning solution does not entirely represent all the milking system CIP cycles.

Therefore, a more generalized mechanism-based mathematical model is needed to describe the milking system CIP. In earlier study (Chapter 5), a set of deposit removal rate-driven models was developed for the milking system CIP process by using a surface evaluation simulator of stainless steel straight pipe specimens to find out the raw milk deposit removal process during the CIP cycles of warm water rinse, alkaline wash and acid wash using EO water as the wash solutions.

Results showed the significant amount of residual removed during the warm water rinse cycle

(more than 90% on average) and the existence of a fast deposit removal rate for the loosely bound deposit and a slow deposit removal rate for the tightly bound deposit. The deposit removal behaved differently during different CIP cycles, and the models were adjusted accordingly.

However, this kind of information is missing for one step CIP process with blended EO water.

Therefore, this study was undertaken to conduct a follow-on study of previous work and investigate the deposit removal process during the optimal blended EO water one-step CIP using the simulator and develop weight-based mechanistic models. ATP bioluminescence method was similarly used as an alternative evaluation at the sampling time points and scanning electron microscopy (SEM) was used to understand the residual deposit morphology.

177

6.3 Materials and methods

6.3.1 Stainless steel surface evaluation by using the simulator

The surface evaluation simulator was used for this study (fig. 5-1). Five 152.4 mm in length stainless steel straight pipe specimens were used as test sections. The simulator had a 1.5 m glass visualization pipe, a 4.6 m stainless steel entrance length pipe, followed by three upstream located specimens (specimen #1, 2, and 3) with 1.1 m in length of recovery section separating them, then after a return elbow there were two additional specimens (specimen #4 and

5) located at the downstream locations. All other essential milking system components such as a claw with liners, solution sink and receiver jar, vacuum system and milk pump were also included as specifically stated in the materials and methods section in Chapter 5.

6.3.2 Experimental procedure

A total of 38 L of raw milk, freshly collected from the Penn State Dairy Barn, was used to soil the simulator. Five-time soiling was used to acquire the initial deposit on the inner surface of the specimen, i.e., one fifth (7.6 L) of raw milk followed by a 10 min air dry, repeated five times (Material and method section in Chapter 5). The milk was not recirculated in the simulator to avoid churning, cream separation, and other physio-chemical changes in the milk properties.

After freshly soiled, the specimens were placed in an incubator (Model Symphony, VWR international, Radnor, PA) and dried for 7 hr, and cooled in the desiccator (Model Dry-keeper

AutoA-3B, Sanplatec Corp., Osaka, Japan) overnight then weighed using a high precision balance

(Model XP504, Mettler-Toledo, LLC., Columbus, OH) with a weighing range of 0 – 520 g and a readability of +/- 0.0001 g. The overnight moisture-balanced deposit mass was the initial deposit mass on the specimen inner surfaces for the soiling process. In a similar manner, the remaining deposit mass after the one-step wash cycle was measured, after 24 hr of the completion of the one-step CIP process.

178

The five specimens were soiled on day 1 and washed using the blended EO water on day

2. Based on the description from Chapter 5, the warm water rinse cycle was conducted for a total time duration of 30 s. The optimal condition of the blended EO water, for the one-step CIP, was set as a 60% of acidic EO water in the blended EO water solution with a starting temperature of

59°C (Wang et al., 2015). The total blended EO water one-step wash cycle was 17 min (1020 s).

The sampling time points during the blended EO water one-step wash cycle were set at 0, 30,

150, 400, and 1020 s for both the upstream and downstream locations. A nondimensionalized evaluation of the residual deposit mass after the one-step wash cycle (mg) per initial deposited mass after the soiling (mg) per unit inner surface area (m2) of the specimen takes the difference of the specimen inner surface area and the initial soiling deposit weight into account.

6.3.3 Mathematical model development

The assumptions made during the model development process were similar to those presented in Chapter 5. The cleaning solution, highly turbulent, assured that the nearest (to the inlet) upstream specimen was placed after the fully developed flow. The distance of the specimen was independent of the deposit removal rate even with the existence of the return elbow. During the blended EO water one-step wash, the wash solution temperature dropped from 59.2±0.44˚C to

41.3±0.57˚C, causing a 0.8% fluid density change and a dynamic viscosity change from about

0.467×10-3 to 0.653×10-3 Pa•s. Correspondingly, the Reynolds number of the wash solution dropped from 7×105 at the start of the wash cycle to 5×105 at the completion of the wash cycle.

But this did not alter the model development process significantly, given the wash solution, despite being slightly more viscous at the end of the wash cycle, still maintained in the highly turbulent state and it also reflects the real-world operation. Additionally, the wash solution was highly diluted, even with the entrainment of the deposits from the inner surface of the specimens; deposit removed at the end of blended EO water one-step wash cycle was at most 0.06 mg/mg/m2/L. Therefore, there was no to minimal impact on the solubility of the deposit in the 179 wash solution. Moreover, due to the complexity of the redeposited deposit amount determination, the deposit removal and its redeposition at the downstream locations were modeled using a lumped formulation.

These assumptions simplified the mechanism-based mathematical analysis of deposit removal. A unified mathematical model development process was used for both the upstream (#1,

2 and 3) and downstream specimens (#4 and 5). The generalized hypothesis proposed here was the coexistence of a fast deposit removal process of the loosely bound deposit and a slow deposit removal process of the tightly bound deposit. The hypothesis was verified based on experimental data.

The mass deposit at any time t in the CIP is designated by m(t) and the subscript denotes the cycle of the CIP process as shown in equation 1; wherein, mTOT(t) is the total mass at time t which comprises the deposit removed during the warm water rinse (WAT), blended EO water one-step wash (ONECIP), and the small amount of unremoved residual deposit (RES) after completion of the CIP process.

mTOT(t) = mWAT(t) + mONECIP(t) + mRES(t) (1)

In equation 1,

 mTOT(t) is the total remaining deposit at any time t in the CIP process, 0 s ≤ t ≤ 1020 s,

 mWAT(t) is the deposit removed during the warm water rinse cycle, 0 s ≤ t ≤ 30 s,

 mONECIP(t) is the deposit removed during the blended EO water one-step wash, 30 s ≤ t

≤ 1050 s, and

 mRES(t) is the residual deposit amount that could not be removed even after the

completion of the blended EO water one-step CIP.

180

Based on the assumptions above, a unified overall first order deposit removal rate equation of remaining deposit mass was proposed for the blended EO water one-step CIP as given in equation 2:

푑푚 (푡) 푇푂푇 = −푘 × 푚 (푡) (2) 푑푡 푇푂푇

where, constant k is the deposit removal rate specific to each CIP cycle.

6.3.3.1 Warm water rinse cycle

A two-term first order deposit removal rate model for the warm water rinse cycle developed in Chapter 5 was used in this study as well. The first term represented the fast deposit removal of the loosely bound deposit and the second term represented the slow deposit removal contributing simultaneously of tightly bound deposits throughout the entire warm water rinse cycle (fig. 6-1). The loosely and tightly bound deposit removal rate equations were expressed as follows:

푑푚 (푡) 푊퐴푇,퐹 = −푘 × 푚 (푡) (3) 푑푡 푊퐴푇,퐹 푊퐴푇,퐹

푑푚 (푡) 푊퐴푇,푆 = −푘 × 푚 (푡) (4) 푑푡 푊퐴푇,푆 푊퐴푇,푆

In equation 3, 푚푊퐴푇,퐹(푡) was the loosely bounded deposit mass at time t at the phase of a fast deposit removal rate and 푘푊퐴푇,퐹 represented the fast deposit removal rate constant. Similarly in equation 4, 푚푊퐴푇,푆(푡) was the tightly bounded deposit mass at time t at the phase of a slow removal rate and 푘푊퐴푇,푆 represented the slow deposit removal rate constant.

The integrated loosely and tightly bound deposit removal models were expressed as:

181

(−푘푊퐴푇,퐹×푡) 푚푊퐴푇,퐹(푡) = 푚푊퐴푇,퐹(푡 = 0) × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (5)

(−푘푊퐴푇,푆×푡) 푚푊퐴푇,푆(푡) = 푚푊퐴푇,푆(푡 = 0) × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (6)

Figure 6-1. Illustration of the proposed two-term deposit removal model during the warm water

rinse cycle, a fast deposit removal rate for the loosely bound deposit from 0 to t1 and near

minimal declining contribution for t>t1, and a simultaneous slow deposit removal rate for

the tightly bound deposit (0 – 30 s).

182

The warm water rinse cycle consisted of a simultaneous occurrence of the fast deposit removal of the loosely bound deposit and the slow deposit removal of the tightly bound deposit

(fig. 6-1), was expressed in equation 7 in total:

푚푊퐴푇(푡) = 푚푊퐴푇,퐹(푡) + 푚푊퐴푇,푆(푡), 0 푠 ≤ 푡 ≤ 30 푠 (7)

6.3.3.2 Blended EO water one-step wash

The deposit removal process behaved similarly during the blended EO water one-step wash cycle as to the warm water rinse cycle. The entire wash cycle was a two-term first order kinetic removal process, with the simultaneous existence of a fast deposit removal of the loosely bound deposit and a slow deposit removal of the tightly bound deposit as informed from the experimental data (eqns. 8 and 9).

푑푚 (푡) 푂푁퐸퐶퐼푃,퐹 = −푘 × 푚 (푡) (8) 푑푡 푂푁퐸퐶퐼푃,퐹 푂푁퐸퐶퐼푃,퐹

푑푚 (푡) 푂푁퐸퐶퐼푃,푆 = −푘 × 푚 (푡) (9) 푑푡 푂푁퐸퐶퐼푃,푆 푂푁퐸퐶퐼푃,푆

The loosely bound deposit was removed at a fast deposit removal rate, represented by

푘푂푁퐸퐶퐼푃,퐹 in equation 8 and the tightly bound deposit was removed at a slow deposit removal rate, represented by 푘푂푁퐸퐶퐼푃,푆 in equation 9 (fig. 6-2). The contribution of the loosely bound deposit occurred from the beginning of the wash cycle to around the time of t2, and then remained minimal till the end of the wash cycle. The removal of the tightly bound deposit was undertaken along the entire wash cycle albeit at a much slower deposit removal rate. Upon integration, the loosely and tightly bound deposit removal models during the blended EO water one-step wash cycle could be expressed as follows:

183

(−푘푂푁퐸퐶퐼푃,퐹×푡) (10) 푚푂푁퐸퐶퐼푃,퐹(푡) = 푚푊퐴푇,퐹(푡 = 30) × 푒 , 30 푠 ≤ 푡 ≤ 1050 푠

(−푘푂푁퐸퐶퐼푃,푆×푡) (11) 푚푂푁퐸퐶퐼푃,푆(푡) = 푚푊퐴푇,푆(푡 = 30) × 푒 , 30 푠 ≤ 푡 ≤ 1050 푠

Figure 6-2. Illustration of the proposed two-term deposit removal model during the blended EO

water one-step wash, a fast deposit removal rate for the loosely bound deposit from 30 to

t2 and near minimal declining contribution for t>t2,, and a simultaneous slow deposit

removal rate for the tightly bound deposit (30 – 1050 s).

Where 푚푊퐴푇,퐹(푡 = 30) and 푚푊퐴푇,푆(푡 = 30) were the milk deposit on the specimen inner surface for the remaining loosely bound deposit (at fast deposit removal rate) after 30 s of warm water rinse and the remaining tightly bound deposit (at slow deposit removal rate) after 30 s of warm water rinse, respectively. The total remaining deposit after 30 s of warm water rinse was the total initial milk deposit on the specimen inner surface for the blended EO water one-step

184 wash cycle to start with. Therefore, the mathematical model for total deposit removal during the blended EO water wash was expressed in equation 12:

푚푂푁퐸퐶퐼푃(푡) = 푚푂푁퐸퐶퐼푃,퐹(푡) + 푚푂푁퐸퐶퐼푃,푆(푡), 30 푠 ≤ 푡 ≤ 1050 푠 (12)

6.3.3.3 Residual deposit expression

Similarly to the models established in Chapter 5, it was proposed that at the completion of CIP cycle a residual deposit would remain, expressed as:

mRES(t) = mONECIP(t)–mONECIP(t=1050) = mTOT(t)–mWAT(t=30)–mONECIP(t=1050) (13)

The coefficients of the deposit removal rates during the warm water rinse and blended

EO water one-step wash cycles (kWAT,F,kWAT,S, kONECIP,F and kONECIP,S), were determined from experimental data. Based on the hypothesis of the blended EO water one-step wash cycle of the fast deposit removal rate and slow deposit removal rate, the validations of the developed mathematical models were conducted during the fast deposit removal phase – the developed mathematical models were further validated at sampling time point of 100 s.

6.3.4 Alternative ATP bioluminescence method

In addition to the mass measurement of the residual deposit, ATP bioluminescence was used as an alternative evaluation. The inner surfaces of the specimens, after weighed, were swabbed using ATP swabs and the relative light unit (RLU) readings were recorded as an indication of the presence of soil.

6.3.5 Qualitative evaluation of the specimen inner surface morphology

Specimens were prepared for the qualitative evaluation of the inner surface morphology after the initial soiling and at the end of warm water rinse and the blended EO water one-step 185 wash cycles. The morphology was observed using scanning electron microscopy (NanoSEM 630,

Nanolab Technologies, Milpitas, CA) and the post processing of the images was done using the

ImageJ software (National Institutes of Health (NIH), USA).

6.3.6 Statistical analysis

Three replications were conducted for each sampling time point. Statistical analysis was performed using Minitab (Version 16.2, Minitab Inc, State College, PA). The significant differences in mean values were determined using Tukey’s method at the 95% confidence interval.

6.4 Results and discussions

This study used a set of stainless steel straight pipes as specimens as a follow-on study of the raw milk deposit removal during the blended EO water one-step wash and proposed mechanism-based deposit removal rate mathematical models. The deposit removal rate model during the blended EO water one-step wash cycle was developed based on the nondimensionalized deposit mass on the specimen inner surfaces, which is given in equation 12.

6.4.1 Statistical comparisons

Initial statistical analyses of the nondimensionalized deposit mass were conducted for all the upstream (#1, 2, and 3) and downstream located specimens (#4 and 5) at all CIP sampling time points. No significant differences were observed among the three upstream (P>0.05) or the downstream located specimens (P>0.05). Therefore, all of three upstream located specimens (#1,

2, and 3) were combined and treated as “upstream locations” and the downstream located specimens (#4 and 5) (P>0.05) were combined and treated as “downstream locations”. During the blended EO water one-step wash cycle, for both the upstream and downstream locations, there

186 were no significant differences for the last two sampling points, i.e. the 400 and 1020 s sampling time points (P>0.05).

6.4.2 Mathematical model development

6.4.2.1 Warm water rinse cycle

The parameter values for equations 5 and 6 for the upstream locations were determined through the experimental data. The resulting equations were:

(−푡/0.4) 푚푊퐴푇_푈푝푠푡푟푒푎푚,퐹(푡) = 50.91 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (14)

(−푡/30.0) 푚푊퐴푇_푈푝푠푡푟푒푎푚,푆(푡) = 6.73 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (15)

Similarly, the equations for downstream locations were:

(−푡/0.9) 푚푊퐴푇_퐷표푤푛푠푡푟푒푎푚,퐹(푡) = 50.41 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (16)

(−푡/55.0) 푚푊퐴푇_퐷표푤푛푠푡푟푒푎푚,푆(푡) = 6.54 × 푒 , 0 푠 ≤ 푡 ≤ 30 푠 (17)

The term of characteristic time was used for analyzing purpose; it represented the time needed to remove 63.2% ((1-e-1)×100%) of the deposit, and therefore a shorter characteristic time indicated a much easily removed deposit with a fast deposit removal rate; and a longer characteristic time indicated a more difficult deposit removal process with a slow deposit removal rate. Additionally, the coefficient of the exponential term could be explained as the maximum possible deposit removal amount, given that at the beginning of the cycle (i.e., time 0 of the wash cycle), the coefficient equals to the starting remaining deposit on the inner surface of the specimen, which is the largest amount of the deposit that could be possibly removed during the wash cycle. It was observed that during the warm water rinse cycle the characteristic time was short for both the upstream and downstream locations, which were 0.4 and 0.9 s, respectively.

This indicated a fast deposit rate removal process of the loosely bound deposit removal. The slow deposit removal process could be observed from the relative longer characteristic time, 30.0 and

187

55.0 s for the upstream and downstream locations, respectively. The slow deposit removal process represented a more tightly bound deposit removal of the particulate and granule deposits

(fig. 6-4). The initial deposited soil heavily covered the specimen inner surface (fig. 6-3a) after the soiling process, but became swollen from the water adsorption and scattered from the fluid shear at the completion of the warm water rinse cycle (fig. 6-3b) (Christian, 2003).

Figure 6-3. Typical inner surface morphology for a specimen at the end of initial soiling (a) and

warm water rinse (b) at upstream location; blended EO water one-step wash at upstream

locations.

188

Figure 6-4. Nondimensionalized experimental decrease in deposit weight during the warm water

rinse cycle (symbols) comparison with the proposed model (solid line) for the upstream

(a) and downstream (b) location.

189

6.4.2.2 Blended EO water one-step wash cycle

As suggested by the experimental data, the deposit removal process during the blended

EO water one-step wash followed similar mechanism as of the warm water rinse cycle. Given the specific representative meaning of the maximum possible deposit removal amount, the models during the blended EO water one-step wash cycle were therefore, proposed based on the blended

EO water one-step wash cycle timeline itself (푡푂푁퐸퐶퐼푃) for the purpose of better explanations of the maximum possible deposit removal amount for the loosely and tightly bound deposit at both the upstream and downstream locations; i.e., 푡푂푁퐸퐶퐼푃(0 푠) = 푡(30 푠) and 푡푂푁퐸퐶퐼푃(1020 푠) =

푡(1050 푠). For the upstream locations, based on the experimental data, the maximum possible deposit removal amount of the loosely bound deposit was calculated as 1.05 mg/mg/m2 with a characteristic time of 64.0 s (eqn. 18); and the maximum possible deposit removal amount of the tightly bound deposit was calculated as 1.44 mg/mg/m2 with a characteristic time of 1140.0 s

(eqn. 19). Clearly, the blended EO water one-step wash cycle started with a much lower amount of possibly removable deposit as compared to the warm water rinse cycle, and the characteristic time for both the loosely and tightly bound deposit were much longer in comparison. However, the loosely bound deposit was still relatively easier removed with the blended EO water solution, and the corresponding characteristic time is shorter (64.0 s) as compared to that of the more tightly bound deposit (1140.0 s). It was observed that, at the sampling time point of 400 s of the wash cycle, which was within less than half of the total wash time (1020 s), the deposit was removed about 65% of the starting deposit amount, i.e., the maximum possible deposit removal amount (fig. 6-5a). The additional 620 s of the remaining wash time, however, only removed an additional about 10% of deposit.

190

Figure 6-5. Nondimensionalized experimental decrease in deposit weight during the CIP process

(symbols) comparison with the proposed model (solid line) for the upstream (a) and

downstream (b) location.

191

It was observed from the SEM images that, at the end of the warm water rinse cycle, the specimen inner surface was still covered heavily with the swollen deposit (fig. 6-3b). However, at the end of the blended EO water one-step wash cycle, the specimen inner surface was cleaner; there was still some remaining deposit on the specimen inner surface, but as suggested by the image (fig. 6-3c), the deposit covered only about 20% of the viewing area. The unremoved deposit might possibly come from the denatured protein, because proteins are alkaline soluble, but only slightly soluble with acid solution (Grasshoff, 1999). Therefore, the unremoved deposit might come from the leftover protein, which partially denatured from the relatively low pH of the blended EO water solution. The deposit on the inner surface of the specimen was 0.8±0.47 mg as indicated from the weight measurement, and the RLU readings were 30 on average, which were way below the cutoff RLU readings of 1,000 for a clean stainless steel surface based on the ATP manufacturer.

(−푡푂푁퐸퐶퐼푃/64.0) 푚푂푁퐸퐶퐼푃_푈푝푠푡푟푒푎푚,퐹(푡푂푁퐸퐶퐼푃) = 1.05 × 푒 , (

0 푠 ≤ 푡푂푁퐸퐶퐼푃 ≤ 1020 푠 (18)

(−푡푂푁퐸퐶퐼푃/1140.0) 푚푂푁퐸퐶퐼푃_푈푝푠푡푟푒푎푚,푆(푡푂푁퐸퐶퐼푃) = 1.44 × 푒 , (

0 푠 ≤ 푡푂푁퐸퐶퐼푃 ≤ 1020 푠 (19)

For the downstream specimens, the maximum possible deposit removal amount for the loosely bound deposit was calculated as 2.07 mg/mg/m2 with a characteristic time of 42.4 s. For the tightly bound deposit, the maximum possible deposit removal amount was calculated as 1.72 mg/mg/m2 with a characteristic time of 1769.9 s (eqns. 20 and 21, respectively). The loosely bound deposit characteristic time was slightly shorter than the corresponding characteristic time of the upstream specimens; only because it started out with a relative higher maximum possible deposit removal amount, i.e., abundance in the loosely and tightly bound deposit to be removed.

The characteristic time of the tightly bound deposit was even longer, indicating a difficulty in

192 removing these tightly bound particulate and granule deposits from the specimen inner surface.

Similar to upstream specimens, about 65% of the starting deposit was removed within 400 s of the wash cycle time, and the additional 620 s wash time only yielded an additional about 10% deposit removal (fig. 6-5b).

It was also observed a higher coverage of the deposit on the specimen inner surface at the completion of the blended EO water one-step wash at the downstream locations. The deposit covered relative more area of the viewing area, about 65% (fig. 6-3d). This was partially due to the previously stated denaturation of the protein; in addition, it was possible to hypothesize that due to the relatively further distance from the inlet and the reduced wash solution temperature, there was a portion of the suspended soil redeposited on the specimen inner surfaces located at the downstream locations. The deposit mass of the downstream specimens at the completion of the wash cycle was 0.9±0.14 mg on average, which was slightly higher than that of the upstream specimens, with an average RLU reading of 60. The inner surface of the downstream located specimens, however, could still be considered as clean based on the cutoff RLU reading.

(−푡푂푁퐸퐶퐼푃/42.4) 푚푂푁퐸퐶퐼푃_퐷표푤푛푠푡푟푒푎푚,퐹(푡푂푁퐸퐶퐼푃) = 2.07 × 푒 , (20) 0 푠 ≤ 푡푂푁퐸퐶퐼푃 ≤ 1020 푠

(−푡푂푁퐸퐶퐼푃/1769.9) 푚푂푁퐸퐶퐼푃_퐷표푤푛푠푡푟푒푎푚,푆(푡푂푁퐸퐶퐼푃) = 1.72 × 푒 , (21) 0 푠 ≤ 푡푂푁퐸퐶퐼푃 ≤ 1020 푠

6.4.2.3 Overall model summary

Our previous study demonstrated that more than 90% of the initial soil was removed at the end of the 30 s warm water rinse for both the upstream and downstream locations (Results and Discussions section in Chapter 5). It was observed from this study that, the blended EO water one-step wash cycle contributed to an additional about 4% (3.3% and 4.9 % for the upstream and downstream locations, respectively) of deposit removal. In comparison, during the alkaline and

193 acidic EO water CIP process, the alkaline wash cycle removed an additional 4% of deposit

(majorly organic deposit including protein and lipid) and the acid wash cycle removed an additional 1% of deposit (mainly inorganic deposit of minerals) (Chapter 5). Therefore, it is reasonable to assume that the blended EO water one-step wash cycle did not achieve a CIP performance as good as that from using the alkaline and acidic EO water for the CIP.

The characteristic times during the warm water rinse cycle were shorter when compared to the blended EO water one-step wash, which suggested that as the CIP progressed, it became harder and harder to remove the remaining deposit and it needs more energy input and wash time to achieve a satisfactory cleaning performance.

The root mean square errors (RMSE) of the developed models for the blended EO water one-step wash were 0.12 and 0.05 mg/mg/m2 for the upstream and downstream locations, respectively, which were acceptable as compared to the magnitude of the deposit amount (the starting nondimensionalized deposit weight of around 60 mg/mg/m2). Additionally, the percent error differences (eq. 22) for the blended EO water one-step wash were 8.07% and 2.24% for the upstream and downstream locations, respectively, which were both considered acceptable.

Percent Error Difference = (RMSE / Experimental deposit mean value)*100% (22)

The validation of the developed models was conducted at 100 s for the blended EO water one-step wash cycle. The term percent variation was used for comparison. It considers the differences between the values from the model prediction with the experimental acquired data as compared to the experimental data, and expressed in eqn. 23.

(푀표푑푒푙 푝푟푒푑푖푐푡푖표푛 −퐸푥푝푒푟푖푚푒푛푡푎푙 푑푎푡푎) Percent Variation = × 100% (23) 퐸푥푝푒푟푖푚푒푛푡푎푙 푑푎푡푎

The percent variation for the upstream and downstream locations was 0.41% and 1.89%, respectively. It is clear that the model predicted values match with the experimental data well with low percent variation, indicating that the developed models were successfully validated at 194 the selected time point. For both the upstream and downstream locations, the developed models over-predicted slightly; potential reasons might come from the limited time sampling points during the model development process causing the potential imperfection, and this could be solved by increasing the number of time sampling points. However, given the overall percent variation of only 1.15%, it is concluded that the developed models were successfully validated at the selected time points with acceptable percent variation.

6.4.3 ATP bioluminescence validation

ATP bioluminescence was used to evaluate the performance of the blended EO water one-step wash. Logarithm of RLU+1 was used for comparison and to avoid the indeterminate value for log of zero. Lower RLUs were preferred to represent a relative cleaner surface and ideally a RLU of 0 was desirable to represent a thoroughly cleaned, no residual soil surface.

The ATP bioluminescence was used at each sampling time point during the blended EO water one-step wash for both the upstream and downstream locations (fig. 6-6). For the upstream specimens, the RLU readings decreased almost 1 log after 30 s of the blended EO water one-step wash, and more than 2.5 log after 150 s of the blended EO water one-step wash. At the end of blended EO water one-step wash cycle, the RLU reading was not the ideal 0, but as stated above, from the cutoff RLU recommendations (RLU below 1,000 for the stainless steel material), the specimen inner surface could be considered as clean (Wang et al., 2013). However, it was observed that the RLU reading at the end of blended EO water one-step wash cycle was more than 1 log higher than that of the previously optimized 10 s warm water rinse/3 min alkaline wash/6 min acid wash CIP (Chapter 5); this finding was not unexpected, as indicated by the deposit mass measurement stated above and informed by the specimen inner surface morphology from the SEM images.

195

Figure 6-6. Nondimensionalized deposit mass and log(RLU+1) of at the sampling time points of

blended EO water one-step CIP and comparison with that of at the end of the previously

optimized 10 s warm water rinse/3 min alkaline wash/6 min acid wash CIP result.

6.5 Conclusions

In this study, a two-term exponential decaying model was developed for the blended EO water one-step wash cycle to describe the fast-removed loosely bound bulk deposits and the slow- removed tightly bound particulate and granule deposits. The developed model had a small root mean square error and percent error difference for both the upstream (0.12 mg/mg/m2 and 2.24%, respectively) and downstream (0.05 mg/mg/m2 and 8.07%, respectively) locations, which suggests the deposit removal model could be considered as acceptable. Experimental data indicated that at the end of the blended EO water one-step wash cycle, an average of an additional

4% deposit was removed. The average residual deposit weight on the specimen inner surfaces

196 was 0.61±0.25 and 0.98±0.11 mg/mg/m2 for the upstream and downstream locations, respectively. The average RLU readings were 30 and 60 for the upstream and downstream locations, respectively, at the end of the blended EO water one-step wash cycle. As indicated from the cutoff RLU reading of 1,000 of stainless steel material, the specimen inner surface at the completion of the blended EO water one-step wash could be considered as clean. Therefore, it is concluded that the blended EO water could be considered as an alternative candidate for one-step

CIP.

6.6 Acknowledgements

Funding for this project was provided in-part by a USDA Special Research Grant (No.

6.7 References

ASABE Standards. (2011). AD5707:2007: Milking machine installations – Construction and

performance. St. Joseph, Mich.: ASABE.

Bird, M.R., & Fryer, P.J. (1991). An experimental study of the cleaning of surfaces fouled by

whey proteins. Food Bioprod Proc., 69,13-21.

Burton, H. (1968). Deposits from whole milk in heat treatment plant-a review and discussion. J

Dairy Res., 35,317–330.

Christian, G. K. (2003). Cleaning of carbohydrate and dairy protein deposits. PhD diss.

Birmingham, UK: University of Birmingham, Department of Chemical Engineering.

197

Dev,S.R.S., Demicri, A., Graves, R.E., & Puri, V.M. (2014). Optimization and modeling of an

electrolyzed oxidizing water based Clean-In-Place technique for farm milking systems

using a pilot-scale milking system. J. Food Eng. 135,1–10.

DPC. (2010). Number 4: Guidelines for installation, cleaning, and sanitizing of large and multiple

receiver parlor milking systems. Richboro, PA.:Dairy Practices Council.

Gallot-Lavallee, T & Lalande, M. (1985). A mechanistic approach of pasteurized milk deposit

cleaning. In Proc. of the 2nd Int. Conf. on Fouling and Cleaning in Food Processing, (pp.

374-394). Madison, WI.

Gillham, C.R., Fryer, P.J., Hasting, A.P.M., & Wilson, D.I. (1999). Cleaning-in-place of whey

protein fouling deposits: mechanisms controlling cleaning. Food Bioprod Proc.,77,127-

136.

Grasshoff, A. (1997). Cleaning of heat treatment equipment. International Dairy Federation (IDF)

monograph, Fouling and cleaning in Heat Exchangers.

Harper, W. J. (1972). Sanitization in dairy food plants. In Food Sanitization, (pp. 130-160).

Westport, CT: The Avi Publ. Co., Inc.

Kulkarni, S. M., Maxcy, R. B., & Arnold, R. G. (1974). Evaluation of soil deposition and removal

processes: an interpretive review. J. Dairy Sci. 58(12), 1922 – 1936.

Kundu, P. J., Cohen, I. M., & Dowling, D. R. (2012). Fluid Mechanics. 5th edition. Elsevier,

Waltham, MA, USA.

Lyster, R. L. J. (1965). The composition of milk deposits in an ultra-high-temperature plant. J.

Dairy Res. 32. 203 – 210.

198

Mohsenin, N. N. (1986). Physical Properties of Plant and Animal Materials: Structure, Physical

Characteristics, and Mechanical Properties. New York.:Gordon and Breach, Science

Publishers.

Monken A. & W. Ingalls. (2002). Milking system cleaning and sanitizing: troubleshooting milk

bacteria counts. In National Mastitis Council Regional Meeting Proceedings, (pp. 55-60).

Kansas City, MI.

Reinemann, D. (1995). System design and performance testing for cleaning milking systems. In

NRAES Conf. Designing a Modern Milking Center: Parlors, Milking Systems,

Management and Economics. Indianapolis, Ind.: NRAES.

Schlussler, H. J. (1976). Kinetics of the cleaning action on hard surfaces. Brauwissenschaft.,

29,263-276.

Wang, X., Dev, S.R.S., Demirci, A., Graves, R.E., & Puri, V.M. (2013). Electrolyzed oxidizing

water for cleaning-in-place of on-farm milking systems performance evaluation and

assessment. Appl. Eng. Agric., 29,717–726.

Wang, X., Demirci, A., Puri, V.M., & Graves, R.E. (2015). Evaluation of blended electrolyzed

oxidizing water-based Cleaning-In-Place technique using a lab scale pilot milking

system. Trans. ASABE. (submitted)

Xin, H., Chen, X., & Ozkan, N. (2004). A mathematical model of the removal of milk protein

deposit. In ECI Conference on Heat Exchanger Fouling and Cleaning: Fundamentals

and Applications, (pp. 158-166). Santa Fe, NM.

199

CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

The increasing trend in the production and consumption of dairy products requires continuing attention to the quality and safety of the dairy products all over the world. The raw material in making the dairy products, milk, is one of the beginning steps that need to be protected from contamination. In the highly automated modern dairy farms, the raw milk, collected from the cow, is usually transported to the bulk tank for storage through a series of milking systems consisting mainly of stainless steel pipelines. Any potential growing microorganisms on the inner surfaces of these pipelines pose threat to the raw milk safety; which is the significance of this study – the cleaning and sanitizing of these milking system pipelines.

Previous studies in our lab had demonstrated the capability of using electrolyzed oxidizing (EO) water as an alternative cleaning-in-place (CIP) approach for the milking system, by testing the milking system related materials and on a lab scale pilot milking system. Therefore, this study examined the performance of using EO water on a commercial dairy farm as a real- world validation. During the four month trial, it was observed that using EO water for the milking system CIP could achieve comparable cleaning effectiveness as compared to the conventional

CIP method. In addition, the operational cost analysis revealed that using EO water CIP was lower than using the conventional CIP by ca. 25% if the EO water generator unit is already in place.

A new CIP protocol was also tested in this study, the one-step cleaning. By combining the alkaline wash cycle and acid wash cycle together as one, the new CIP protocol saved the usage of water, chemicals, energy, and CIP operational time. The goal was to examine the possibility of using a certain blending ratio of alkaline and acidic EO waters as an alternative one-step CIP. An optimization including different percentages of acidic EO water in the blended EO water, 200 cleaning time duration and the starting temperature of the blended EO water solution was conducted to find the optimal condition for the blended EO water one-step CIP. Results showed that the combination of a cleaning time of 17 min, a starting temperature of the blended EO water solution of 59°C, and an acidic EO water percentage of 60% in the blended EO water solution resulted in a 100% RLU reduction percentages for both sampling locations of pipes and elbows.

Similarly, to evaluating the CIP performance and cost of the optimal blended EO water solution, two commercially available one-step CIP products were used as comparison. Based on the results, the optimal blended EO water solution was able to achieve a similar level of cleaning and sanitizing performance when compared to the commercially available one-step CIP chemicals.

Moreover, the cost comparisons revealed an 80% reduction in the operational cost when using the optimal blended EO water solution compared to the commercially available one-step CIP chemicals.

To achieve a better fundamental understanding of the CIP process, a series of studies were undertaken to evaluate the raw milk deposit removal kinetics using the EO water solutions.

A surface evaluation simulator made from mostly stainless steel pipes and elbows were used.

Stainless steel straight pipes were used as specimens and evaluated through raw milk deposit mass measurement. The raw milk deposit on the inner surface of the specimen was evaluated after the soiling process and at several time points during the warm water rinse cycle, alkaline wash cycle and acid wash cycles, which comprised the CIP process. Mathematical models were developed to quantitatively determine the raw milk deposit removal kinetics during each CIP cycle, and results showed a different deposit removal behavior during different CIP cycles. It was observed here again the importance of the warm water rinse conducted right after the milking event is finished; results showed that a 10 s warm water rinse would result in more than 90% of the initial raw milk deposit removal. Two different deposit removal terms were suggested including a fast deposit removal of the loosely bound deposit and a simultaneous slow deposit

201 removal of the tightly bound deposit. Additional evaluation was conducted using the ATP bioluminescence and the CIP process was further shortened as indicated from the developed models. The CIP was shortened by 55% (from 20.5 to 9.17 minutes) based on the developed models to achieve a satisfactory CIP performance for the simulator used in this study.

Additionally, the relationship between the ATP bioluminescence RLU reading and the deposit mass after the initial soiling was studied and a near-linear regression was proposed (an average R2 of 0.911). The developed mathematical models had a good fit with the experimental data and successfully validated at designated time points (an average RMSE of 0.15 mg/mg/m2 and an average percent difference of 2.30%). The shortened CIP process was promising given the potential saving in energy consumption and resource usage.

Subsequently, a follow-on study was conducted for the optimal blended EO water one- step CIP modelling. In a similar manner to the EO alkaline and acidic water-based models, a two- term exponential deposit removal model was developed for the blended EO water one-step wash.

The proposed models, when compared to the experimental data, had an average of 0.08 mg/mg/m2 of root mean square errors and an average of 5.16% of low percentage error differences for the upstream and downstream locations, which could be considered as acceptable for the developed model. Additionally, the developed mathematical models were validated at the one-step wash time of 70 s successfully with a percent variation of 1.15%. Moreover, as indicated from the cutoff RLU reading, the specimen inner surface at the completion of the blended EO water one-step wash was considered as clean.

To summarize, in this study the performance of using alkaline and acidic EO water as alternative of the conventional milking system CIP solutions were validated on a commercial dairy farm scale, and the performance of using an optimized blended EO water as an alternative of the one-step CIP solution was validated on a lab scale pilot milking system; both of which indicated relative lower operational cost once the EO water unit is in place and a comparable CIP

202 performance when compared to the typically used CIP chemicals. It is therefore suggested that

EO water is indeed a promising technology that could possibly applied in vast areas for the milking system. More importantly, in this dissertation, the kinetic models for the milking system

CIP were developed and validated by using a stainless steel surface evaluation simulator. The developed models depicted the raw milk deposit removal process during the warm water rinse cycle, alkaline wash cycle, acid wash cycle, and also the optimized blended EO water one-step wash cycle. The deposit removal behaved differently in different cycles, and the deposit removal contribution during each cycle was also quantified, which could shed light on the potential further

CIP cycle optimization.

Despite the significant progress made in this dissertation, there are still knowledge gaps.

Following are some recommendations for the future studies:

The EO water technology was invented more than a decade ago, but the successful expansion and scale-up marketing still needs effort. There are several possible reasons hindering the process, and overcoming these obstacles would be beneficial for successful marketing. The electrolyzing cell needs to be replaced on a regular basis, for example, the life span of the electrolyzing cell used in this study is 3,000 working hours. The additional charge for this replacement needs to be lowered for a cheaper maintenance fee. Another possible reason is the

EO water generation rate. On larger scale of dairy farms or food processing plants, the EO water solution generation rate (for example, 100 L/hr for the EO water generator used in this study) cannot really meet the requirement in these plants. Therefore, high capacity system is also needed.

For the alkaline and acidic EO water farm trial, longer term evaluation is needed before the application to the commercial market. Moreover, more than one single farm is needed to be included under different environmental conditions such as the supply water properties, the pipeline configurations and so on. During the blended EO water one-step CIP study, more work is 203 needed for the real world application, and similar studies could be conducted on commercial dairy farms and certain adjustments might be necessary due to the differences in lab and on farm.

When scaled up in quantity, the blending ratio needed might also be adjusted accordingly based on the time-temperature profile during the blending process.

There had been concerns on dairy farms of the raw milk contamination coming from the contaminated teat cups contacting the sick cows during the continuous milking process. The blended EO water one-step wash study in this dissertation has the potential to solve this problem.

Nowadays, the typical handling procedure is to rinse the teat cups with warm water after milking the sick cows and then continued to milk other cows (healthy ones); this warm water rinse could of course remove some contamination (as could be similarly deduced from the raw milk deposit mathematical modeling part) but still poses high bacterial contamination potential. If rinsed with the blended EO water solution, the sanitizing performance would be much better, as compared to using only warm water. However, to achieve this, more studies are needed to determine the blended EO water solution blending ratio and temperature along with the rinsing time duration.

Short time duration is particularly desired to decrease the downtime during the milking.

There are more adjustments and improvements that could be made to the mathematical modelling studies as well. The distance effect was not completely verified in the current study given the limited space in lab, but is definitely needed for better understanding of the raw milk deposit removal mechanism including re-deposition specially during recirculated wash cycles.

Only stainless steel straight pipe segments were tested as specimens in the current study, but as stated above, more complex configurations (with more material choices such as glass, plastic, and rubber) exist in the real world scenario, such as different angles of elbows and some potential contraction and expansion connectors.

The flow profile would be disturbed at these configurations and the deposit removal behavior would alter accordingly, which would be an interesting topic to study. Deposit weight 204 measurement was used as the major evaluation during the current study for the modelling process; this could be better improved by either increasing the deposit amount by using other soiling method such as manually layer on the deposit or adding microorganisms into the raw milk; or acquiring more accurate balance to measure the deposit amount (down to 0.01mg or higher) and reduce the weight of the specimens used (typically weighed about 300 g in this study). Moreover, it is importance to investigate other potential deposit evaluation methods other than weight measurement such as the heat flux difference. The deposit was evaluated at only a number of discrete time sampling points in the current study, and this could be further improved by developing a continuous evaluation method, perhaps in situ and without the disassembly of the specimens; that would reveal more realistically of what the deposit removal mechanisms are.

The in-line analysis/detection, if successful, would be of great benefit to the researchers and the farmers. Scanning electron microscopy was used for the inner surface morphology inspection in the current study, but there are definitely a lot more room for improvement. The viewing magnification could be further increased, if needed, and other microscopic methods

(such as optical profilometer for a 3D mapping of the deposit surface) could be also used to examine the surface morphology from other perspectives such as the deposit layer height and weight and exact deposit surface coverage. Additionally, it would be interesting to know the exact residual component in the deposit by investigating their chemical properties (such as FT-IR or

XPS), and potentially identifying the layers/types of deposit proposed in this study.

205

REFERENCES

Abadias, M., Usall, J., Oliveira, M., Alegre, I., & Vinas, I. (2008). Efficacy of neutral electrolyzed

water (NEW) for reducing microbial contamination on minimally-processed vegetables.

Intl. J. Food Microbiol., 123,151-158.

Alameda, J. E., Bravo, R. V., Bailon, M. R., Olea, N. J., & Vaz, D. A. (2011). Fatty soils removal

from hard surfaces in a clean-in-place system. J. Food Proc. Engr., 34(4), 1053–1070.

Alfa-Laval, A.B. (1995). Dairy Handbook. Alfa-Laval, Lund, Sweden.

ASABE Standards. (2008). Milking machine installations – Vocabulary. ANSI/ASABE

AD3918:2007. ASABE, St. Joseph, MI.

ASABE Standards. (2011). AD5707:2007: Milking machine installations – Construction and

performance. St. Joseph, Mich.: ASABE.

Audenaert, K., Monbaliu, S., Deschuyffeleer, N., Maene, P., Vekeman, F., Haesaert, G., Saeger,

D. S., & Eeckhout, M. (2012). Neutralized electrolyzed water efficiently reduces

Fusarium spp. in vitro and on wheat kernels but can trigger deoxynivalenol (DON)

biosynthesis. Food Cont., 23, 515–521.

Augustin, M., Ali-Vehmas, T., & Atroshi, F. (2004). Assessment of enzymatic cleaning agents and

disinfectants against bacterial biofilms. J. Pharmacy & Pharmaceutical Sci., 7, 55–64.

Balasubramanian, S., & Puri, V. M. (2008a). Reduction of fouling during tomato juice

pasteurisation in plate heat exchanger systems using food-grade surface coating. Food

Manufacturing Efficiency, 2, 1–13.

Balasubramanian, S., & Puri, V. M. (2008b). Fouling mitigation during product processing using

a modified plate heat exchanger surface. Transaction of ASABE, 51, 629–639.

206

Balasubramanian, S., & Puri, V. M. (2009a). Reduction of milk fouling in a plate heat exchanger

system using food-grade surface coating. Trans. ASABE, 52, 1603–1610.

Balasubramanian, S., & Puri, V. M. (2009b). Thermal energy savings in pilot-scale plate heat

exchanger system during product processing using modified surfaces. J. Food Engr., 91,

608–611.

Banerjee, S., Yang, R., & Courchene, C. E. (2009). Scanning electron microscopy measurements

of the surface roughness of paper. Ind. Eng. Chem. Res., 48, 4322–4325.

Bansal, R. K. (2005). A Text Book of Fluid Mechanics and Hydraulic Machines, (pp. 286-292) New

Delhi, India.: Laxmi Publications.

Barnes, L. M., Lo, M. F., Adams, M. R., & Chamberlain, A. H. (1999). Effect of milk proteins on

adhesion of bacteria to stainless steel surfaces. Appl. Env. Microbiol., 65, 4543–4548.

Baumann, A. R., Martin, S. E., & Feng, H. (2009). Removal of Listeria monocytogenes biofilms

from stainless steel by use of ultrasound and ozone. J. Food Prot., 72, 1306–1309.

Beecher, C. (2013). Tennessee E. Coli Outbreak Linked to Raw Milk Dairy Food Safety News,

Retrieved from http://www.foodsafetynews.com/2013/11/raw-milk-dairy-linked-to-e-

coli-outbreak-through-tests/#.Uv0yJPldV8F.

Benjamin, S., & Labib, M. (2000). Turbulent two phase flow; a novel method for cleaning and

high grade disinfection of flexible endoscopies and internal channels. Am. J.

Gastroenterology., 95, 2492–2492.

Bernbom, N., Ng, Y. Y., Jørgensen, R. L., Arpanaei, A., Meyer, R. L., Kingshott, P., Vejborg, R.

M., Klemm, P., & Gram, L. (2009). Adhesion of food-borne bacteria to stainless steel is

reduced by food conditioning films. J. Appl.Micro., 106, 1268–1279.

207

Bialka, K. L., Demirci, A., Knabel, S. J., Patterson, P. H., & Puri, V. M. (2004). Efficacy of

electrolyzed oxidizing water for the microbial safety and quality of eggs. Poultry Sci.,

83, 2071–2078.

Bird, M.R., & Fryer, P.J. (1991). An experimental study of the cleaning of surfaces fouled by

whey proteins. Food Bioprod Proc., 69,13-21.

Blel, W., Bénézech, T., Legentilhomme, P., Legrand, J., & Le Gentil-Lelièvre, C. (2007). Effect of

flow arrangement on the removal of Bacillus spores from stainless steel equipment

surfaces during a Cleaning In Place procedure. Chem. Engr. Sci., 62(14), 3798–3808.

Blel, W., Legentilhomme, P., Le Gentil-Lelièvre, C., Faille, C., Legrand, J., & Bénézech, T.

(2010). Cleanability study of complex geometries: Interaction between B. cereus spores

and the different flow eddies scales. Biochem. Engr. J., 49(1), 40–51.

Borucki, M. K., Peppin, J. D., White, D., Loge, F., & Call, D. R. (2003). Variation in biofilm

formation among strains of Listeria monocytogenes. Appl. Env. Microbiol., 69, 7336–

7342.

Bott, T. R. (1995). Fouling of heat exchangers. New York: Elsevier.

Bowser, T. J., Weckler, P. R., Patil, K. N., & DeWitt, C. (2005). Design and testing of a low-cost,

pilot-scale batch gasifier for food processing byproducts. Appl. Engr. Agric., 21(5), 901–

906.

Buck, J. W., Iersel, M. W., Oetting, R. D., & Hung, Y. C. (2002).. In vitro fungicidal activity of

acidic electrolyzed oxidizing water. Plant Disease, 86(3), 278–281

Burgess, S. A., Lindsay, D., & Flint, S. H. (2010). Thermophilic bacilli and their importance in

dairy processing. Int. J. Food Micro., 144, 215–225.

208

Burgess, S. A., Brooks, J. D., Rakonjac, J., Walker, K. M., & Flint, S. H. (2009). The formation of

spores in biofilms of Anoxybacillus flavithermus. J. Appl. Micro., 107, 1012–1018.

Burton, H. (1968). Section G. Deposits from whole milk in heat treatment plant—a review and

discussion. J. Dairy Res., 35, 317.

Cais, S. D., & Pikul, J. (2008). Evaluation of steel surface cleanliness level in dairies using the

bioluminescence method. Bull Vet Inst Pulawy., 52, 625–629.

Carrascosa, C., Saavedra, P., Millan, R., Jaber, J. R., Perez, E., Grau, R., Raposo, A., Mauricio,

C., & Sanjuan, E. (2012). Monitoring of cleanliness and disinfection in dairies:

comparison of traditional microbiological and ATP bioluminescence methods. Food

Cont., 28, 367 – 373.

Caubet, R., Pedarros-Caubet, F., Chu, M., Freye, E., de Belém Rodrigues, M., Moreau, J. M., &

Ellison, W. J. (2004). A radio frequency electric current enhances antibiotic efficacy

against bacterial biofilms. Antimicrobial Agents and Chemotherapy, 48, 4662–4664.

Center for Dairy Excellence. (2012). Overview of Pennsylvania’s Dairy Industry. Retrieved from

http://centerfordairyexcellence.org/pennsylvania-dairy-industry-overview/

Chen, F.-C., & Godwin, S. L. (2006). Comparison of a rapid ATP bioluminescence assay and

standard plate count methods for assessing microbial contamination of consumers’

refrigerators. J. Food Prot., 69, 2543–2548.

Chmielewski, R., & Frank, J. (2003). Biofilm formation and control in food processing facilities.

Comp. Reviews Food Sci. & Food Safety, 2, 22–32

Choi, W., Jun, S., Nguyen, L. T., Rungraeng, N., Yi, H., Balasubramanian, S., Puri, V. M., & Lee,

J. (2013). 3-D milk fouling modeling of plate heat exchangers with different surface

finishes using computational fluid dynamics codes. J. Food Proc. Engr., 36, 439–449.

209

Christian, G. K. (2003). Cleaning of carbohydrate and dairy protein deposits. PhD diss.

Birmingham, UK: University of Birmingham, Department of Chemical Engineering.

Christian, G. K., & Fryer, P. J. (2006). The effect of pulsing cleaning chemicals on the cleaning of

whey protein deposits. Food & Bioproducts Proc., 84, 320–328.

Craigen, B., Dashiff, A., & Kadouri, D. E. (2011). The use of commercially available alpha-

amylase compounds to inhibit and remove Staphylococcus aureus biofilms. The Open

Microbiology J., 5, 21–31.

Dat, N. M., Hamanaka, D., Tanaka, F., & Uchino, T. (2012). Control of milk pH reduces biofilm

formation of Bacillus licheniformis and Lactobacillus paracasei on stainless steel. Food

Cont., 23, 215–220.

Davey, M. E., Caiazza, N. C., & O’Toole, G. A. (2003). Rhamnolipid surfactant production

affects biofilm architecture in Pseudomonas aeruginosa PAO1. J.Bacteriology, 185,

1027–1036.

Del Pozo, J. L., Rouse, M. S., Mandrekar, J. N., Sampedro, M. F., Steckelberg, J. M., & Patel, R.

(2009). Effect of electrical current on the activities of antimicrobial agents against

Pseudomonas aeruginosa, Staphylococcus aureus, and Staphylococcus epidermidis

biofilms. Antimicrobial Agents and Chemotherapy, 53, 35–40.

Dev, S. R. S., Demirci, A., Graves, R., & Puri, V. (2014). Optimization and modeling of an

electrolyzed oxidizing water based clean-in-place technique for farm milking systems

using a pilot- scale milking system. J. Food Engr., 135, 1–10.

Di Bonaventura, G., Piccolomini, R., Paludi, D., D’Orio, V., Vergara, A., Conter, M., & Ianieri, A.

(2008). Influence of temperature on biofilm formation by Listeria monocytogenes on

various food-contact surfaces: relationship with motility and cell surface hydrophobicity.

J. Appl. Micro., 104, 1552–1561.

210

Donlan, R. M. (2002). Biofilms: microbial life on surfaces. Emerging Infectious Diseases Journal

8, 881–890.

Dosti, B., Guzel-Seydim, Z., & Greene, A. K. (2005). Effectiveness of ozone, heat and chlorine

for destroying common food spoilage bacteria in synthetic media and biofilms. Intl. J.

Dairy Tech., 58, 19–24.

DPC. (2006). Guideline for effective installation, cleaning, and sanitizing of tie barn milking

systems. Dairy Practices Council Publication Number: 102.

DPC (2010). Guidelines for installation, cleaning, and sanitizing of large and multiple receiver

parlor milking systems. Dairy Practices Council Publication No 4.

Dürr, H., & Graßhoff, A. (1999). Milk heat exchanger cleaning: modelling of deposit removal.

Food Bioprod. Proc., 77, 114 – 118.

Fabrizio, K. A., & Cutter, C. N. (2004). Comparison of electrolyzed oxidizing water with other

antimicrobial interventions to reduce pathogens on fresh pork. Meat Sci., 68, 463–468.

Fabrizio, K. A., & Cutter, C. N. (2005). Application of electrolyzed oxidizing water to reduce

Listeria monocytogenes on ready-to-eat meats. Meat Sci., 71, 327–333.

Fasenko, G. M., Christopher, E. E. O., & McMullen, L. M. (2009). Spraying hatching eggs with

electrolyzed oxidizing water reduces eggshell microbial load without compromising

broiler production parameters. Poultry Sci., 88, 1121–1127.

FAOSTAT. (2012). Total production of cow milk, whole fresh cow. Rome, Italy: Food and

Agriculture Organization of the United Nations. Retrieved from

http://faostat3.fao.org/faostat-gateway/go/to/browse/Q/QL/E.

Ferreira, C., Pereira, A. M., Melo, L. F., & Simões, M. (2010). Advances in industrial biofilm

control with micro-nanotechnology. In: Current Research, Technology and Education

211

Topics in Applied Microbiology and Microbial Biotechnology. (A. Mendez-Vilas), 2nd ed,

pp. 845–854. Formatex Research Center, Spain.

FDA. (2012). The dangers of raw milk: unpasteurized milk can pose a serious health risk,

Retrieved from

http://www.fda.gov/downloads/Food/FoodborneIllnessContaminants/UCM239493.

Flint, S., Bremer, P., & Brooks, J. (1997). Biofilms in dairy manufacturing plant‐description,

current concerns and methods of control. Biofouling. J. Bioadh. Biofilm Res., 11, 81–97.

Flint, S. H., Brooks, J. D., & Bremer, P. J. (2000). Properties of the stainless steel substrate,

influencing the adhesion of thermo-resistant streptococci. J. Food Engr., 43, 235–242.

Fornalik, M. (2008). Detecting biofouling in food processing systems, Retrieved from

http://www.photonics.com/Article.aspx?AID=32596.

Föste, H., Schöler, M., Majschak, J.-P., Augustin, W., & Scholl, S. (2013). Modeling and

validation of the mechanism of pulsed flow cleaning. Heat Transfer Engr., 34(8-9), 753–

760.

Fratamico, P. M., Annous, B. A., & Cunther IV, N. W. (Eds.) (2009). Biofilms in the food and

beverage industries, 1st ed. Woodhead Publishing Limited, Cambridge, UK.

Gallot-Lavallee, T & Lalande, M. (1985). A mechanistic approach of pasteurized milk deposit

cleaning. In Proc. of the 2nd Int. Conf. on Fouling and Cleaning in Food Processing, (pp.

374-394). Madison, WI.

Gaspar-Rolle, M. N. P. (1991). Attachment of bacteria of Teflon ad buna-N-rubber gasket materials.

Ph.D. Dissertation. Blacksburg, Va.: Virginia Polytechnic Institute and State University.

212

GEA Farm Technologies. (2012). Retrieved from http://www.gea-

farmtechnologies.com/us//en/bu/farm_services/hygiene/equipment_facility/cip/cip_disinfec

tion/zinicin/default.aspx.

Georgiadis, M. C., & Macchietto, S. (2000). Dynamic modelling and simulation of plate heat

exchangers under milk fouling. Chemical Engr. Sci. 55, 1605–1619.

Gillham, C. R., Fryer, P. J., Hasting, A. P. M., & Wilson, D. I. (1999). Cleaning-in-place of whey

protein fouling deposits. Food & Bioprod. Proc., 77, 127–136.

Grasshoff, A. (1999). Fundamental trials to clean UHT tubular modules. Retrieved from

https://getinfo.de/app/Fundamental-trials-to-clean-UHT-tubular-

modules/id/BLCP:CN030907072.

Guentzel, J. L., Lam, K. L., Callan, M. A, Emmons, S. A., & Dunham, V. L. (2008). Reduction of

bacteria on spinach, lettuce, and surfaces in food service areas using neutral electrolyzed

oxidizing water. Food Microbiol., 25, 36–41.

Guentzel, J. L., Lam, K. L., Callan, M. A, Emmons, S. A., & Dunham, V. L. (2010). Postharvest

management of gray mold and brown rot on surfaces of peaches and grapes using

electrolyzed oxidizing water. Intl. J. Food Microbiol., 143, 4–60.

Guentzel, J. L., Lam, K. L., Callan, M. A, Emmons, S. A., & Dunham, V. L. (2011). Evaluation of

electrolyzed oxidizing water for phytotoxic effects and pre-harvest management of gray

mold disease on strawberry plants. Crop Prot., 30, 1274–1279.

Harper, W. J. (1972). Sanitization in dairy food plants. In Food Sanitization, (pp. 130-160).

Westport, CT: The Avi Publ. Co., Inc.

HHS/FDA. (2007). Guidance for industry: Dairy farms, bulk milk transporters, bulk milk transfer

stations and fluid milk processors: food security preventive measures guidance.

Washington, D.C.: HHS/FDA. 213

Hricova, D., Stephan, R., & Zweifel, C. (2008). Electrolyzed water and its application in the food

industry. J. Food Prot., 71(9), 1934–1047.

Hsu, S. Y., & Kao, H. Y. (2004). Effects of storage conditions on chemical and physical properties

of electrolyzed oxidizing water. J. Food Eng., 65,465–471.

Hsu, S. Y. (2005). Effects of flow rate, temperature and salt concentration on chemical and

physical properties of electrolyzed oxidizing water. J. Food Eng., 66,171 – 176.

Huang, Y.-R., Hung, Y.-C., Hsu, S.-Y., Huang, Y.-W., & Hwang, D.-F. (2008). Application of

electrolyzed water in the food industry. Food Cont., 19, 329–345.

Huang, Y. C., Hsieh, H. S., Lin, S. Y., Lin, S. J., Hung, Y. C., & Hwang, D. F. (2006). Application

of electrolyzed oxidizing water on the reduction of bacterial contamination for seafood.

Food Cont., 17, 987 – 993.

Huang, Y. C., Tilly, P., & Kim, C. (2010). Efficacy of electrolyzed oxidizing (EO) water and

chlorinated water for inactivation of Escherichia Coli O157:H7 on strawberries and

broccoli. J. Food Qua., 33, 559–577.

Iversen, C., Lane, M., & Forsythe, S. J. (2004). The growth profile, thermotolerance and biofilm

formation of Enterobacter sakazakii grown in infant formula milk. Letters in Appl.

Micro., 38, 378–382.

Jayasooriya, S. D., Bhandari, B. R., Torley, P., & D’Arcy, B. R. (2004). Effect of high power

ultrasound waves on properties of meat: a review. Intl. J. Food Prop., 7, 301–319.

Jeurnink, T.J.M. & Brinkman, D.W. (1994). The cleaning of heat exchangers and evaporators

after processing milk or whey. Int. Dairy J., 4, 347 – 368.

Jones, G.M. (2009). Cleaning and saniziting milking equipment. Retrieved from

http://pubs.ext.vt.edu/404/404-400/404-400.html.

214

Jullien, C., Bénézech, T., Carpentier, B., Lebret, V., & Faille, C. (2003). Identification of surface

characteristics relevant to the hygienic status of stainless steel for the food industry. J.

Food Engr., 56(1), 77–87.

Jun, S., & Puri, V. M. (2004). Fouling performance of plate heat exchangers using a 2D dynamic

model. NABEC 2004 Paper No: 04-0016. ASAE, St. Joseph, MI.

Kallioinen, M., & Mänttäri, M. (2011). Influence of ultrasonic treatment on various membrane

materials: a review. Separation Sci. Tech., 46, 1388–1395.

Kaplan, J. B., Ragunath, C., Ramasubbu, N., & Fine, D. H. (2003). Detachment of Actinobacillus

actinomycetemcomitans biofilm cells by an endogenous beta-hexosaminidase activity. J.

Bacteriol., 185, 4693–4698.

Kaplan, J. B., Ragunath, C., Velliyagounder, K., Fine, D. H., & Ramasubbu, N. (2004).

Enzymatic detachment of Staphylococcus epidermidis biofilms. Antimicrob. Agents

Chemother., 48, 2633–2636.

Klahre, J., & Flemming, H. (2000). Monitoring of biofouling in papermill process waters. Water

Res., 34, 3657–3665.

Knight, G. C., Nicol, R. S., & McMeekin, T. A. (2004). Temperature step changes: a novel

approach to control biofilms of Streptococcus thermophilus in a pilot plant-scale cheese-

milk pasteurisation plant. Intl. J. Food Microbiol., 93, 305–18.

Kraigsley, A., Ronney, P. D., & Finkel, S. E. (2014). Hydrodynamic influences on biofilm

formation and growth. Retrieved from

http://carambola.usc.edu/research/biophysics/biofilms4web.html.

Kulkarni, S. M., Maxcy, R. B., & Arnold, R. G. (1974). Evaluation of soil deposition and removal

processes: an interpretive review. J. Dairy Sci. 58(12), 1922 – 1936.

215

Kumar, C. G., & Anand, S. K. (1998). Significance of microbial biofilms in food industry: a

review. Intl. J. Food Microbiol., 42, 9–27.

Kundu, P. J., Cohen, I. M., & Dowling, D. R. (2012). Fluid Mechanics. 5th edition. Elsevier,

Waltham, MA, USA.

Larson, E. L., Aiello, A. E., Gomez-Duarte, C., Lin, S. X., Lee, L., Della-Latta, P., & Lindhardt,

C. (2003). Bioluminescence ATP monitoring as a surrogate marker for microbial load on

hands and surfaces in the home. Food Microbiol., 20, 735–739.

Latorre, A. A., Van Kessel, J. S., Karns, J. S., Zurakowski, M. J., Pradhan, A. K., Boor, K. J.,

Jayarao, B. M., Houser, B. A., Daugherty, C. S., & Schukken, Y. H. (2010). Biofilm in

milking equipment on a dairy farm as a potential source of bulk tank milk contamination

with Listeria monocytogenes. J. Dairy Sci., 93(6), 2792–802. doi:10.3168/jds.2009-2717

Le Gentil, C., Sylla, Y., & Faille, C. (2010). Bacterial re-contamination of surfaces of food

processing lines during cleaning in place procedures. J. Food Engr., 96(1), 37–42.

Lelièvre, C., Legentillhomme, P., Gaucher, C., Legrand, J., Faille, C., & Bénézech, T. (2002).

Cleaning in place: effect of local wall shear stress variation on bacterial removal from

stainless steel equipment. Chem. Engr., Sci., 57(8), 1287 – 1297.

Leon, M. B., & Albrecht, J. A. (2007). Comparison of adenosine triphosphate (ATP)

bioluminescence and aerobic plate counts (APC) on plastic cutting boards. J. Foodserv.,

18, 145–152.

Liao, L. B., Chen, W. M., & Xiao, X. M. (2007). The generation and inactivation mechanism of

oxidation–reduction potential of electrolyzed oxidizing water. J. Food Engr., 78(4),

1326–1332. doi:10.1016/j.jfoodeng.2006.01.004

216

Liu, W., Chtistian, G. K., Zhang, Z., & Fryer, P. J. (2006a). Direct measurement of the force

required to disrupt and remove fouling deposits of whey protein concentrate. Int. Dairy

J., 16, 164–172

Liu, W., Zhang, Z., Chtistian, G. K., & Fryer, P. J. (2003). Direct measurement of the forces

required to disrupt and remove fouling deposits. Heat Exchanger Fouling and Cleaning:

Fundamentals and Applications. Vol. RP1, Article 25.

Liu, W., Zhang, Z., & Fryer, P. J. (2006b). Identification and modeling of different removal

modes in the cleaning of a model food deposit. Chemical Engr. Sci., 61, 7528–7534.

Lyster, R. L. J. (1965). The composition of milk deposits in an ultra-high-temperature plant. J.

Dairy Res. 32. 203 – 210.

Mafu, A. A., Plumety, C., Deschênes, L., & Goulet, J. (2011). Adhesion of pathogenic bacteria to

food contact surfaces: influence of pH of culture. Intl. J. Microbiol., 2011, 972494.

Marchand, S., De Block, J., De Jonghe, V., Coorevits, A., Heyndrickx, M., & Herman, L. (2012).

Biofilm formation in milk production and processing environments; influence on milk

quality and safety. Comp. Rev.Food Sci. Food Safety, 11, 133–147.

Marth, E. H. (1988). Fundamentals of dairy chemistry. Springer.

McLandsborough, L., Rodriguez, A., Pérez-Conesa, D., & Weiss, J. (2006). Biofilms: at the

interface between biophysics and microbiology. Food Biophysics, 1, 94–114.

Mittelman, M. W. (1998). Structure and functional characteristics of bacterial biofilms in fluid

processing operations. J. Dairy Sci., 81, 2760–2764.

Mohsenin, N. N. (1986). Physical Properties of Plant and Animal Materials: Structure, Physical

Characteristics, and Mechanical Properties. New York.:Gordon and Breach, Science

Publishers.

217

Monken A. & W. Ingalls. (2002). Milking system cleaning and sanitizing: troubleshooting milk

bacteria counts. In National Mastitis Council Regional Meeting Proceedings, (pp. 55-60).

Kansas City, MI.

Monroe, D. (2007). Looking for chinks in the armor of bacterial biofilms. PLOS Biology, 5,

2458–2461.

Nagamatsu, Y., Chen, K. K., Tajima, K., Kakigawa, H., & Kozono, Y. (2002). Durability of

bactericidal activity in electrolyzed neutral water by storage. Dental Materials J., 21(2):

93 – 104.

O’Toole, G., Kaplan, H. B., & Kolter, R. (2000). Biofilm formation as microbial development.

Annual Rev. Microbiol., 54, 49–79.

O’Toole, G. A., & Kolter, R. (1998). Flagellar and twitching motility are necessary for

Pseudomonas aeruginosa biofilm development. Mol. Microbiol., 30, 295–304.

Ogden, K. (1993). Practical experiences of hygiene control using ATP-bioluminescence. J. Inst.

Brew., 99, 389–393.

Okull, D. O., & Laborde, L. F. (2004). Activity of electrolyzed oxidizing water against penicilium

expansum in suspension and on wounded apples. Food Microbiol. Safety, 69(1), 23–27.

Okull, D. O., Demirci, A., Rosenberger, D., & Laborde, L. F. (2006). Susceptibility of Penicillium

expansum spores to sodium hypochlotite, electrolyzed oxidizing water, and chlorine

dioxide solutions modified with nonionic surfactants. J. Food Prot., 69(8), 1944–1948.

Oliveira, K., Oliveira, T., Teixeira, P., Azeredo, J., Henriques, M., & Oliveira, R. (2006).

Comparison of the adhesion ability of different Salmonella enteritidis serotypes to

materials used in kitchens. J. Food Prot., 69, 2352–2356.

218

Olofsson, A.-C., Hermansson, M., & Elwing, H. (2003). N-acetyl-L-cysteine affects growth,

extracellular polysaccharide production, and bacterial biofilm formation on solid

surfaces. Appl. Env. Microbiol., 69, 4814–4822.

Oulahal, N., Martial-Gros, A., Bonneau, M., & Blum, L. J. (2004). Combined effect of chelating

agents and ultrasound on biofilm removal from stainless steel surfaces. Application to

Escherichia coli milk and Staphylococcus aureus milk biofilms. Biofilms, 1, 65–73.

Ozden, H. O., & Puri, V. M. (2010). Computational analysis of fouling by low energy surfaces. J.

Food Engr., 99, 250–256.

Ozer, N. P., & Demirci, A. (2006). Electrolyzed oxidizing water for decontamination of raw

salmon inoculated with Escherichia coli O157:H7 and Listeria monocytogenes Scott A

and response surface modeling. J. Food Eng., 72, 234–241.

Parkar, S. G., Flint, S. H., & Brooks, J. D. (2004). Evaluation of the effect of cleaning regimes on

biofilms of thermophilic bacilli on stainless steel. J. Appl. Microbiol., 96, 110–116.

Parr, K. (2013). New exacta one-Step CIP detergent. AgroChem Farm Dairy Prod. Retrieved from

http://www.agrocheminc.com/images/brochures/exacta_press_release.pdf.

Percival, S. L., Knapp, J. S., Wales, D. S., & Edyvean, R. G. J. (1999). The effect of turbulent

flow and surface roughness on biofilm formation in drinking water. J. Ind. Microbiol.

Biotechnol., 22, 152–159.

Pereira, A., Mendes, J., & Melo, L. F. (2008). Using nanovibrations to monitor biofouling.

Biotechnol.Bioeng., 99, 1407–1415.

Poortinga, A. T., Smit, J., van der Mei, H. C., & Busscher, H. J. (2001). Electric field induced

desorption of bacteria from a conditioning film covered substratum. Biotech. & Bioengr.,

76, 395–399.

219

Powell, C. J., & Jablonski, A. (2009). Surface sensitivity of X-ray photoelectron spectroscopy.

Nuc. Instr. Met. in Physics Res., 601, 54–65.

Powell, G. L., Smyrl, N. R., Williams, D. M., & Barber, T. E. (2009). Surface inspection using

fourier transform infrared sprectroscopy. U.S. Department of Energy under contract DE-

AC05-84OR21400.

Reinemann, D. (1995). System design and performance testing for cleaning milking systems. In

Natural Resource, Agriculture, and Engineering Service Conference Designing a Modern

Milking Center: Parlors, Milking Systems, Management and Economic. NRAES

Conference, Rochester, NY.

Reinemann, D. (2002). Application of cleaning and cooling principles to robotic milking.

DeLaval Inaug. Symp. Kansas City MO.

Reinemann, D., & Ruegg, P. (2000). An investigation of ATP bioluminescence and quantitative

bulk cultures to assess cleanliness of milking machines. ASAE Annual Intl. Meeting. July.

Paper No. 003009, 1–8.

Ren, T. T., & Su, Y. C. (2006). Effects of electrolyzed oxidizing water treatment on reducing

Vibrio parahaemolyticus and Vibrio vulnificus in raw oysters. J. Food Prot., 69(8), 1829–

1834.

Rijnaarts, H. H., Norde, W., Bouwer, E. J., Lyklema, J., & Zehnder, A. J. (1993). Bacterial

adhesion under static and dynamic conditions. Appl. & Environmental Micro., 59, 3255–

3265.

Rosmaninho, R., & Melo, L. F. (2006). Calcium phosphate deposition from simulated milk

ultrafiltrate on different stainless steel-based surfaces. Int. Dairy. J., 16, 81–87.

220

Rosmaninho, R., & Melo, L. F. (2007). Effect of proteins on calcium phosphate deposition in

turbulent flow as a function of surface properties. Experimental Thermal & Fluid Sci.,

32(2), 375–386.

Rosmaninho, R., & Melo, L. F. (2008). Protein–calcium phosphate interactions in fouling of

modified stainless-steel surfaces by simulated milk. Intl. Dairy J., 18(1), 72–80.

Rosmaninho, R., Santos, O., Nylander, T., Paulsson, M., Beuf, M., Benezech, T., Yiantsios, S.,

Andritsos, N., Karabelas, A., Rizzo, G., Muller-Steinhagen, H., & Melo, L. F. (2007).

Modified stainless steel surfaces targeted to reduce fouling – Evaluation of fouling by

milk components. J. Food Engr., 80(4), 1176–1187.

Rousseau, D., Sonwai, S., & Khan, R. (2010). Microscale surface roughening of chocolate view

with optical profilometry. J. Am. Oil. Chem. Soc., 87, 1127–1136.

Ryu, J.-H., & Beuchat, L. R. (2005). Biofilm formation and sporulation by Bacillus cereus on a

stainless steel surface and subsequent resistance of vegetative cells and spores to

chlorine, chlorine dioxide, and a peroxyacetic acid-based sanitizer. J. Food Prot., 68,

2614–2622.

Sakurai, Y., Nakatsu, M., Sato, Y., & Sato, K. (2003). Endoscope contamination from HBV- and

HCV- positive patients and evaluation of a cleaning/disinfection method using strongly

acidic electrolyzed water. Digest. Endosc., 15, 19–24.

Samkutty, P. J., Gough, R. H., Adkinson, R. W., & McGrew, P. (2001). Rapid assessment of the

bacteriological quality of raw milk using ATP bioluminescence. J. Food Prot.., 64, 208–

212.

Sanford, S. (2003). Heating water on dairy farms. University of Wisconsin – Cooperative Extension

Publication A3784-2. Madison, Wis.: University of Wisconsin.

221

Sauer, K., Camper, A. K., Ehrlich, G. D., Costerton, J. W., & Davies, D. G. (2002). Pseudomonas

aeruginosa displays multiple phenotypes during development as a biofilm. J. Bacteriol.,

184, 1140–1154.

Schlussler, H. J. (1976). Kinetics of the cleaning action on hard surfaces. Brauwissenschaft.,

29,263-276.

Schöler, M., Föste, H., Helbig, M., Gottwald, A., Friedrichs, J., Werner, C., Augustin, W., Scholl,

S., & Majschak, J.-P. (2012). Local analysis of cleaning mechanisms in CIP processes.

Food Bioprod. Proc., 90(4), 858–866.

Serraino, A., Veronese, G., Alonso, S., Matera, R., Lugoboni, B., & Giacometti, F. (2010).

Bactericidal activity of electrolyzed oxidizing water on food processing surfaces. Intl. J.

Food. Sci., 2(22), 222–228.

Shaheen, R., Svensson, B., Andersson, M. A., Christiansson, A., & Salkinoja-Salonen, M. (2010).

Persistence strategies of Bacillus cereus spores isolated from dairy silo tanks. Food

Micro. 27, 347–355.

Sharma, M., & Anand, S. K. (2002). Characterization of constitutive microflora of biofilms in

dairy processing lines. Food Micro., 19, 627–636.

Sharma, R. R., & Demirci, A. (2003). Treatment of Escherichia coli O 157: H7 inoculated alfalfa

seeds and sprouts with electrolyzed oxidizing water. Int. J. Food Microbiol., 86(3), 231-

237.

Sheng, X., Ting, Y. P., & Pehkonen, S. O. (2008). The influence of ionic strength, nutrients and

pH on bacterial adhesion to metals. J. Coll. Interf. Sci., 321, 256–264.

Shirtliff, M. E., Bargmeyer, A., & Camper, A. K. (2005). Assessment of the ability of the

bioelectric effect to eliminate mixed-species biofilms. Appl. Env.. Microbiol., 71, 6379–

6382. 222

Silva, S., Teixeira, P., Oliveira, R., & Azeredo, J. (2008). Adhesion to and viability of Listeria

monocytogenes on food contact surfaces. J. Food Prot., 71, 1379–1385.

Simões, M., Simões, L. C., Machado, I., Pereira, M. O., & Vieira, M. J. (2006). Control of flow-

generated biofilms with surfactants. Food Bioprod. Proc., 84, 338–345.

Sinde, E., & Carballo, J. (2000). Attachment of Salmonella spp. and Listeria monocytogenes to

stainless steel, rubber and polytetrafluorethylene: the influence of free energy and the

effect of commercial sanitizers. Food Microbiol., 17, 439–447.

Speers, J. G. S., & Gilmour, A. (1985). The influence of milk and milk components on the

attachment of bacteria to farm dairy equipment surfaces. J. Appl. Bacteriol., 59, 325–332.

Stepanović, S., Ćirković, I., Mijač, V., & Švabić-Vlahović, M. (2003). Influence of the incubation

temperature, atmosphere and dynamic conditions on biofilm formation by Salmonella

spp. Food Microbiol., 20, 339–343.

Stepanović, S., Cirković, I., Ranin, L., & Svabić-Vlahović, M. (2004). Biofilm formation by

Salmonella spp. and Listeria monocytogenes on plastic surface. Lett.Appl. Microbiol., 38,

428–432.

Stevenson, S. M. L., Cook, S. R., Bach, S. J., & McAllister, T. A. (2004). Effects of water source,

dilution, storage, and bacterial and fecal loads on the efficacy of electrolyzed oxidizing

water for the control of Escherichia coli O157:H7. J. Food Prot., 67(7), 1377–1383.

Stoodley, P., Cargo, R., Rupp, C. J., Wilson, S., & Klapper, I. (2002a). Biofilm material properties

as related to shear-induced deformation and detachment phenomena. J. Ind. Microbiol.

Biotechnol., 29, 361–367.

Stoodley, P., Sauer, K., Davies, D. G., & Costerton, J. W. (2002b). Biofilms as complex

differentiated communities. Ann. Rev. Microbiol., 56, 187–209.

223

Tamine, A. Y., Almeida, K. E., & Oliveira, M. N. (2008). Acidification rates of probiotic bacteria

in Minas frescal cheesy whey. Food Sci. Technol., 41(2), 311–316

Teixeira, R., Pereira, A., Mendes, J., & Melo, L. F. (2014). Identifying the nature of fouling layers

by online monitoring of the propagation of vibrations along the deposition surface. Heat

Transfer Engr., 35, 251–257.

Tissier, J. P., & Lalande, M. (1986). Experimental device and methods for studying milk deposit

formation on heat exchange surfaces. Biotechnol. Prog., 2(4), 218–29.

Toyoda, I., & Fryer, P. J. (1997). A computational model for reaction and mass transfer in fouling

from whey protein solutions: In Fouling Mitigation of Industrial Heat- Exchange

Equipment, 589-597. C. B. Panchal, Begell House, New York, NY.

Turner, D. E., Daugherity, E. K., Altier, C., & Maurer, K. J. (2010). Efficacy and limitations of an

ATP-based monitoring system. J. Ame. Asso. Lab.Animal Sci., 49, 190–195.

Unnerstad, H., & Bannerman, E. (1996). Prolonged contamination of a dairy with Listeria

monocytogenes. Netherlands Milk Dairy J., 50, 493–499.

USDA. (2012). USDA Economic Research Service - Dairy Data, Retrieved from

http://www.ers.usda.gov/data-products/dairy-data.aspx#.Uv0x2vldV8E.

Vilar, M. J., Rodríguez-Otero, J. L., Diéguez, F. J., Sanjuán, M. L., & Yus, E. (2008). Application

of ATP bioluminescence for evaluation of surface cleanliness of milking equipment. Intl.

J. Food Microbiol., 125, 357–361.

Visser, J. (1995). Particle adhesion and removal: a review. Particu. Sci. Technol., 13, 169–196.

Vlkova, H., & Babak, V. (2008). Biofilms and hygiene on dairy farms and in the dairy industry:

sanitation chemical products and their effectiveness on biofilms–a Review. Czech J. Food

Sci., 26, 309–323.

224

Walker, S. P. (2003). Cleaning milking systems using electrolyzed oxidizing water. MS Thesis.

University Park, Pa.: Pennsylvania State University, Department of Agricultural and

Biological Engineering.

Walker, S. P., Demirci, A., Graves, R. E., Spencer, S. B., & Roberts, R. F. (2005a). Response surface

modeling for cleaning and disinfecting materials used in milking systems with electrolyzed

oxidizing water. Int. J. Dairy Technol., 58(2), 65-73.

Walker, S. P., Demirci, A., Graves, R. E., Spencer, S. B., & Roberts, R. F. (2005b). Cleaning

milking systems using electrolyzed oxidizing water. Trans. ASAE, 48(5), 1827-1833.

Wallhäußer, E., Sayed, A., Nöbel, S., Hussein, M. A., Hinrichs, J., & Becker, T. (2013).

Determination of cleaning end of dairy protein fouling using an online system combining

ultrasonic and classification methods. Food Bioproc. Technol., 7, 506–515.

Wang, X., Dev, S.R.S., Demirci, A., Graves, R.E., & Puri, V.M. (2013). Electrolyzed oxidizing

water for cleaning-in-place of on-farm milking systems performance evaluation and

assessment. Appl. Eng. Agric., 29,717–726.

Wang, X., Demirci, A., Puri, V.M., & Graves, R.E. (2015). Evaluation of blended electrolyzed

oxidizing water-based Cleaning-In-Place technique using a lab scale pilot milking

system. Transactions of ASABE. (submitted)

Weltman, A., Longenberger, A. H., Moll, M., Johnson, L., Martin, J., & Beaudoin, A. (2013).

Notes from the field: recurrent outbreak of Campylobacter jejuni infections associated

with a raw milk dairy — Pennsylvania, April–May 2013. Retrieved from

http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6234a4.htm.

Wijman, J. G. E., de Leeuw, P. P. L. A., Moezelaar, R., Zwietering, M. H., & Abee, T. (2007). Air-

liquid interface biofilms of Bacillus cereus: formation, sporulation, and dispersion. Appl.

Env.. Microbiol., 73, 1481–1488.

225

Wilson, D. I., Fryer, P. J., & Hasting, A. P. M. (2002). Fouling, cleaning and disinfection in food

processing. Conference proceedings. UK: Department of Chemical Engineering

University of Cambridge.

Wirtanen, G. (1995). Biofilm formation and its elimination from food processing equipment.

Technical Research Center of Finland.

Wong, N. P. (1988). Fundamentals of Dairy Chemistry, Van Nostrand Reinhold Company, New

York, NY.

Xiong, K., Liu, H.-J., Liu, R., & Li, L.-T. (2010). Differences in fungicidal efficiency against

Aspergillus flavus for neutralized and acidic electrolyzed oxidizing waters. Int. J. Food

Microbiol., 137(1), 67–75.

Xin, H., Chen, X., & Ozkan, N. (2004). A mathematical model of the removal of milk protein

deposit. In ECI Conference on Heat Exchanger Fouling and Cleaning: Fundamentals

and Applications, (pp. 158-166). Santa Fe, NM.Xiong, K., Liu, H. J., Liu, R., & Li, L. T.

(2010). Differences in fungicidal efficiency against Aspergillus flavus for neutralized and

acidic electrolyzed oxidizing waters. J. Food Microbiol., 137, 67–75

Zhao, Q. (2004). Effect of surface free energy of graded NI–P–PTFE coatings on bacterial

adhesion. Surf. Coatings Technol., 185, 199–204.

Zhao, Q., & Liu, Y. (2006). Modification of stainless steel surfaces by electroless Ni-P and small

amount of PTFE to minimize bacterial adhesion. J. Food Engr., 72, 266–272.

Zhao, Q., Liu, Y., & Wang, C. (2005a). Development and evaluation of electroless Ag-PTFE

composite coatings with anti-microbial and anti-corrosion properties. Appl. Surf. Sci.,

252, 1620–1627.

226

Zhao, Q., Liu, Y., Wang, C., Wang, S., & Müller-Steinhagen, H. (2005b). Effect of surface free

energy on the adhesion of biofouling and crystalline fouling. Chem. Engr. Sci., 60, 4858–

4865.

227

APPENDIX A1

ORIGINAL DATA DURINGT THE ALKALINE AND ACID WASH CIP

PROCESS FOR THE UPSTREAM LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) warm water rinse alkaline wash acid wash

Time, s--> 0 5 20 30 60 120 210 630 660 720 810 1230 L1#1 58.2567 5.1967 3.0616 2.2241 4.2094 1.4536 0.8967 0.1826 0.5655 0.3807 0.5805 0.0990 L1#2 58.0079 7.0330 2.5278 2.1891 2.3498 1.5248 0.1083 0.8388 0.6237 0.4364 0.3800 0.3983 L1#3 58.2576 7.1714 2.3060 2.1613 1.5237 3.7725 0.6985 0.1832 0.4297 0.4098 0.7501 0.1572 L2#1 57.1750 4.6096 1.2240 3.1673 1.5001 1.2004 0.8792 0.1014 0.8951 0.5743 0.2862 0.3394 L2#2 57.1750 5.2383 2.4578 3.2436 1.1776 1.8220 0.2094 0.7779 0.4026 0.2936 0.1821 0.0000 L2#3 57.4748 3.2493 3.9845 3.1329 2.4979 1.5681 0.3512 0.7729 0.0000 0.6403 0.4163 0.3471 L3#1 57.8925 4.5665 3.4978 1.6482 0.6732 1.2796 1.0460 0.3554 0.7346 0.6873 0.0000 0.0371 L3#2 56.8835 6.2526 3.1108 1.5142 3.4896 0.7342 0.5775 0.4130 0.3856 0.5199 0.0000 0.4077 L3#3 57.6825 6.9421 3.1976 3.1589 2.8999 1.1932 0.2524 0.9731 0.4748 0.5118 0.5949 0.3143

Nondimensionalized mass per unit contact surface area – Validation (mg/mg/m2)

warm water rinse alkaline wash acid wash

Time, s--> 2 135 135 L1#1 6.8954 0.9719 0.3278 L1#2 8.1869 1.0383 0.3210 L1#3 4.6981 1.5636 0.4432 L2#1 5.2287 1.3071 0.3642 L2#2 6.6483 1.0075 0.0912 L2#3 8.7272 0.7239 0.3833 L3#1 5.5491 0.9403 0.3269 L3#2 5.7576 0.6505 0.6098 L3#3 8.0409 1.3118 0.3600

228

Actual deposit mass (mg) warm water rinse alkaline wash acid wash

Time, s--> 0 5 20 30 60 120 210 630 660 720 810 1230 L1#1 53.8 9.0 2.1 2.2 2.6 2.0 1.2 0.2 0.6 0.4 0.6 0.1 L1#2 44.8 7.5 2.5 2.2 1.7 1.6 0.1 0.8 0.5 0.4 0.4 0.4 L1#3 84.7 8.1 1.9 2.2 1.6 5.2 1.0 0.2 0.5 0.6 0.6 0.2 L2#1 63.1 5.6 1.3 2.6 1.7 1.0 1.3 0.1 0.8 0.6 0.3 0.4 L2#2 58.6 4.7 1.9 3.9 1.3 2.1 0.3 0.8 0.5 0.4 0.2 0.0 L2#3 60.1 6.3 4.9 3.6 2.4 1.8 0.4 1.0 0.0 0.6 0.6 0.3 L3#1 64.4 5.1 4.3 1.2 1.6 1.8 1.4 0.4 0.6 0.6 0.0 0.1 L3#2 56.2 11.2 7.7 4.3 3.4 1.3 1.6 1.1 0.8 0.8 0.0 0.3 L3#3 72.0 8.3 3.8 3.3 2.6 1.4 0.2 1.1 0.5 0.7 0.6 0.3

Actual deposit mass – Validation (mg)

Time, s--> warm water rinse alkaline wash acid wash 2 135 135

L1#1 7.8 0.8 0.4 L1#2 6.3 0.9 0.2 L1#3 6.0 1.0 0.6 L2#1 5.0 1.2 0.3 L2#2 6.0 0.8 0.1 L2#3 13.5 0.6 0.4 L3#1 5.8 0.8 0.5 L3#2 10.9 1.5 0.5 L3#3 6.8 1.1 0.5

229

APPENDIX A2

ORIGINAL DATA DURINGT THE ALKALINE AND ACID WASH CIP

PROCESS FOR THE DOWNSTREAM LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) warm water rinse alkaline wash acid wash

Time, s--> 0 10 30 60 210 630 660 720 810 1230 L4#1 57.7917 3.8961 5.2869 4.3831 0.5452 0.9198 1.3352 0.1907 0.3760 0.3853 L4#2 57.5765 4.3991 3.9603 1.8459 1.3459 2.5628 0.9257 1.7395 0.4122 0.4336 L4#3 55.9275 7.1371 4.0834 4.2371 2.9734 1.8351 0.6581 0.7754 0.8485 0.3236 L5#1 56.8526 5.4454 3.1170 5.3561 0.7546 1.2319 1.1102 1.0135 0.2491 0.2782 L5#2 57.7607 7.7456 3.2609 1.8571 0.7522 0.1584 1.2735 0.5801 0.3517 0.1289 L5#3 56.3081 4.0758 3.0241 2.4471 1.7941 1.4568 0.8159 0.6975 0.4075 0.5266

Nondimensionalized mass per unit contact surface area – Validation (mg/mg/m2)

warm water rinse alkaline wash acid wash

Time, s--> 2 135 135 L4#1 10.1071 2.7996 0.8027 L4#2 14.0308 2.1230 0.7830 L4#3 10.2427 1.9276 0.4726 L5#1 9.3541 1.8093 0.7124 L5#2 15.9643 1.4393 0.6007 L5#3 12.1810 0.7759 0.6492

Actual deposit mass (mg) warm water rinse alkaline wash acid wash

Time, s--> 0 10 30 60 210 630 660 720 810 1230 L4#1 62.3 4.2 2.9 2.7 0.4 0.6 0.9 0.2 0.4 0.3 L4#2 44.5 3.4 2.6 1.0 1.8 1.7 0.5 1.0 0.3 0.3 L4#3 40.2 5.0 1.9 3.5 3.3 1.6 0.5 0.8 0.8 0.3 L5#1 35.4 3.3 3.5 4.3 0.9 1.0 0.8 1.1 0.2 0.2 L5#2 49.1 6.8 2.3 1.1 1.2 0.2 1.4 0.3 0.3 0.1 L5#3 54.4 3.9 3.5 1.1 1.1 0.7 0.6 0.8 0.2 0.4

230

Actual deposit mass – Validation (mg)

warm water rinse alkaline wash acid wash

Time, s--> 2 135 135 L4#1 7.8 1.4 0.8 L4#2 14.5 2.5 0.7 L4#3 6.6 1.3 0.6 L5#1 9.5 3.4 0.4 L5#2 9.8 1.0 0.9 L5#3 9.0 0.6 0.7

231

APPENDIX A3

ORIGINAL DATA DURINGT THE TIME REDUCTED ALKALINE AND

ACID WASH CIP PROCESS FOR THE UPSTREAM LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) Time, s--> 0 10 190 370 550 L1#1 58.2567 3.0412 0.6055 0.4507 0.1188 L1#2 58.0079 4.2912 1.0303 0.6936 0.0000 L1#3 58.2576 5.5677 0.8467 0.5723 0.3002 L2#1 57.1750 6.1615 0.7337 0.6871 0.5217 L2#2 57.1750 3.8832 0.7841 0.3431 0.2331 L2#3 57.4748 2.3607 0.4134 0.8421 0.0832 L3#1 57.8925 5.3731 0.4031 0.6273 0.1635 L3#2 56.8835 3.4231 0.9057 0.3969 0.5660 L3#3 57.6825 5.7674 0.6464 0.4142 0.2111

Actual deposit mass (mg) Time, s--> 0 10 190 370 550 L1#1 53.8 2.8 0.6 0.3 0.1 L1#2 44.8 3.3 1.1 0.4 0.0 L1#3 84.7 8.0 0.8 0.4 0.3 L2#1 63.1 6.8 1.4 0.7 0.5 L2#2 58.6 4.0 0.9 0.3 0.4 L2#3 60.1 2.4 0.4 0.5 0.1 L3#1 64.4 5.9 0.4 0.4 0.4 L3#2 56.2 3.3 1.1 0.8 0.4 L3#3 72.0 7.3 0.6 0.3 0.2

232

APPENDIX A4

ORIGINAL DATA DURINGT THE TIME REDUCTED ALKALINE AND

ACID WASH CIP PROCESS FOR THE DOWNSTREAM

LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) Time, s--> 0 10 190 370 550 L4#1 57.7917 3.8961 1.5208 1.2071 0.1566 L4#2 57.5765 4.3991 1.4865 1.3352 0.2893 L4#3 55.9275 7.1371 2.0117 0.9881 0.4151 L5#1 56.8526 5.4454 2.5233 0.5856 0.3666 L5#2 57.7607 7.7456 0.9168 0.7831 0.6092 L5#3 56.3081 4.0758 1.3699 0.5841 0.0711

Actual deposit mass (mg) Time, s--> 0 10 190 370 550 L4#1 62.3 4.2 1.1 0.8 0.1 L4#2 44.5 3.4 1.5 0.9 0.2 L4#3 40.2 5.0 2.1 1.1 0.4 L5#1 35.4 3.3 2.0 0.6 0.3 L5#2 49.1 6.8 0.6 0.5 0.5 L5#3 54.4 3.9 1.4 0.5 0.1

233

APPENDIX A5

ORIGINAL DATA DURINGT THE OPTIMAL BLENDED EO WATER

ONE-STEP WASH CIP PROCESS FOR THE UPSTREAM

LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) Time, s--> 0 30 150 400 1020 L1#1 2.2241 1.9766 1.4122 0.5165 0.7104 L1#2 2.1891 2.1871 1.2880 1.2210 0.5258 L1#3 2.1613 1.8278 1.0880 0.8028 0.5770 L2#1 3.1673 2.1728 1.7613 0.9498 0.5515 L2#2 3.2436 1.9955 1.6703 0.5471 0.5327 L2#3 3.1329 1.6594 1.5523 1.2887 0.5935 L3#1 1.6482 1.5469 0.6681 0.7829 0.6854 L3#2 1.5142 1.5443 1.8465 0.6013 0.6975 L3#3 3.1589 1.5507 1.0610 1.3947 0.5963

Nondimensionalized mass per unit contact surface area – Validation– onestep wash (mg/mg/m2)

Time, s--> 70 L1#1 1.6267 L1#2 1.5840 L1#3 1.3675 L2#1 1.8389 L2#2 2.6614 L2#3 1.4222 L3#1 1.8296 L3#2 1.6541 L3#3 1.3051

234

Actual deposit mass (mg) Time, s--> 0 30 150 400 1020 L1#1 2.2 1.6 1.0 0.7 0.6 L1#2 2.2 2.1 1.3 0.8 0.5 L1#3 2.2 2.0 1.1 0.7 0.6 L2#1 2.6 2.2 1.4 0.9 0.6 L2#2 3.9 1.6 1.3 0.6 0.5 L2#3 3.6 1.6 1.2 1.4 0.7 L3#1 1.2 3.3 0.9 1.1 0.7 L3#2 4.3 1.8 1.4 1.8 1.9 L3#3 3.3 2.2 1.8 1.0 0.5

Actual deposit mass – Validation– onestep wash (mg)

Time, s--> 70 L1#1 1.3 L1#2 1.2 L1#3 1.0 L2#1 1.8 L2#2 2.9 L2#3 2.1 L3#1 1.8 L3#2 4.8 L3#3 1.2

235

APPENDIX A6

ORIGINAL DATA DURINGT THE OPTIMAL BLENDED EO WATER

ONE-STEP WASH CIP PROCESS FOR THE DOWNSTREAM

LOCATIONS

Nondimensionalized mass per unit contact surface area (mg/mg/m2) Time, s--> 0 30 150 400 1020 L4#1 5.2869 3.0417 2.0230 0.9887 1.1628 L4#2 3.9603 2.7860 1.5252 1.2915 1.0192 L4#3 4.0834 3.0380 1.4436 1.3789 0.8192 L5#1 3.1170 2.5464 2.3424 1.4978 0.9270 L5#2 3.2609 1.9242 1.5116 1.1075 1.0458 L5#3 3.0241 2.4889 0.9989 1.5353 0.9102

Nondimensionalized mass per unit contact surface area – Validation – onestep wash (mg/mg/m2)

Time, s--> 70 L4#1 2.4455 L4#2 1.9081 L4#3 1.7014 L5#1 2.0580 L5#2 2.5442 L5#3 1.4130

Actual deposit mass (mg) Time, s--> 0 30 150 400 1020 L4#1 2.9 1.9 1.7 1.1 1.0 L4#2 2.6 2.4 1.6 1.3 1.0 L4#3 1.9 2.1 1.2 2.6 0.7 L5#1 3.5 2.7 1.6 0.9 0.9 L5#2 2.3 1.1 1.8 0.9 1.1 L5#3 3.5 1.9 0.8 1.1 0.9

236

Actual deposit mass – Validation – onestep wash (mg)

Time, s--> 70 L4#1 1.9 L4#2 1.6 L4#3 1.1 L5#1 1.3 L5#2 2.6 L5#3 1.1

237

CURRICULUM VITAE

Xinmiao Wang (王鑫淼)

[email protected]

EDUCATION  Ph.D., Agricultural and Biological Engineering, The Pennsylvania State University, University Park, PA. (2015)  M.Sc., Food Science and Engineering, Harbin Institute of Technology, Harbin, China. (2011)  B.S., Food Science and Engineering, Harbin Institute of Technology, Harbin, China. (2010)

RESEARCH EXPERIENCE Research assistant at Penn State University  Modelling of the milk deposit removal during electrolyzed oxidizing (EO) water cleaning-in-place (CIP) (2013 – 2015)  Optimization and evaluation of blended EO water for one-step CIP (2013 – 2014)  Evaluation of EO water for on-farm milking system CIP (2011 – 2012) Research assistant at Harbin Institute of Technology  Inhibition mechanism of Hela using resveratrol, sitosterol, and lupeol (2009 – 2011)  Optimization and evaluation of polysaccharide extraction in pine cone from Pinus koraiensis (2008 – 2009)

WORK EXPERIENCE  Teaching assistant for BE 308 (Engineering Elements of Biochemistry and Microbiology) & BE 468 (Microbiological Engineering), University Park, Penn State University (2013 – 2015)